| , NaNth |
, NaNth
8:30-Noon | | Pre-Conference Seminar A: Introduction to 3D NAND (Pre-Conference Seminars Track) | | Speaker(s):
| | Session Description:
| 4-bits per cell (QLC) technology and the push for 128+ layers in 3D NAND flash continue to drive down the cost per gigabyte of solid-state storage. How do these developments affect the performance and reliability of storage systems? What is different about the NAND manufacturers’ technologies? How does 3D XPoint™ affect the flash market? If you are new to flash memory and want to get the most out of FMS, this seminar will get you up-to-speed quickly with the terms and concepts used in the industry. If you are an experienced storage professional, but are curious about the inner workings of flash and how they affect system design, this introduction will quickly give you the fundamentals of the critical aspects of flash The challenges and opportunities presented by increased storage density in 3D (or “vertical”) NAND flash are significant. A clear understanding of the structure and function of flash cells enables designers to get the most out of their systems. How is the changing role of the FTL (flash translation layer) in data center appliances changing how 3D NAND is being used, tested, and optimized? If more of the native NAND performance is exposed, what do you do with it? What can you expect to encounter when using 3D NAND in extreme environments like automotive, industrial, and IoT? How does the trend toward persistent memory affect flash usage? What are the consequences of the billions of dollars China is investing in 3D NAND manufacturing? This seminar is designed for engineers, managers, and executives who must make immediate decisions on implementing or optimizing 3D flash technology. Reliable, unbiased information is critical for making the right decisions! So don’t proceed (and spend a lot of development money) without a clear understanding of the fundamentals, opportunities, and challenges of this critical technology! This seminar is presented by KnowledgeTek, the world’s leading data storage technology training company. Suggested prerequisites include a technical background and an interest in 3D technology. The class does not assume detailed engineering knowledge of flash memory or semiconductor processing. Course Outline 1. The Fundamentals of 3D NAND ● 3D: What it is and what it isn’t ● Moore’s law ended for 2D NAND flash ● Charge trap flash cells ● Roadmaps for 3D NAND ● The device determines the system 2. Competing Technologies and Architectures for 3D Flash ● Vertical transistor designs ● 3D array architectures ● Accessing data ● SLC to TLC and QLC in 3D ● Comparison to MRAM, ReRAM, and other nonvolatile memories 3. Process, Design, and Test Challenges ● Etch and deposition vs. lithography ● Logical-to-physical translation in 3D ● Testing, characterization, and defects ● Controller partitioning 4. Next Steps in 3D ● Persistent memory ● Parameter setting with machine learning ● Will emerging NVMs replace flash? ● China’s flash initiatives ● Opportunities with 3D XPoint™ (QuantX™) ● A lesson from The Wrath of Khan ● What to watch for at the Summit!
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-Noon | | Pre-Conference Seminar B: NVMe over Fabrics (NVMe-oF) (Pre-Conference Seminars Track) | | Speaker(s):
| | Session Description:
| NVM Express (NVMe) is a specification for connecting non-volatile memory devices such as SSDs to computers via the PCIe bus. The idea is to take advantage of the high speed of PCIe (as compared to disk interfaces such as SAS or SATA), while maintaining full standardization and access to a large ecosystem. It has rapidly become a popular interface supported by almost all major SSD makers. Storage designers and end users alike appreciate its combination of low cost, high speed, widespread support, and standardized approach. NVMe started with a clean slate to build a command set and supporting mechanism that would take advantage of the PCIe transport and SSD storage devices. It is scalable from iPads to enterprise and high performance computing. At the Flash Memory Summit, many keynote presentations and breakout sessions assume a basic understanding of NVMe. Be prepared to take full advantage of the Summit by upgrading your background with this pre-conference presentation. This presentation assumes no specific knowledge of PCIe or NVMe, but does assume a general understanding of computers. Outline of Topics: Introduction to NVMe NVMe Operation and Structures Command and status Door bells Queues Interrupts Data flow Controller Memory Buffer Host Memory Buffer Admin Commands Create/Delete Submission/Completion Queues Set/Get Features DWord parameter passing Passing parameters in memory Identify Namespace Management and Attachment I/O Commands Command List Using RDMA NVMe 1.3 Enhancements New Commands Directive Send/Receive Device Self-test Sanitize NVMe-MI Send/Receive Doorbell Buffer Configuration Virtualization Management New Features Time Stamp Host Controller Thermal Management Non-Operational Power State Configuration New Registers Boot Partition Information Boot Partition Read Select Boot Partition Buffer Location New log pages are covered with the new command or feature NVMe 1.4 Enhancement (if released) IO Determinism
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-Noon | | Pre-Conference Seminar C: Persistent Memory (Pre-Conference Seminars Track) | | Speaker(s):
| | Session Description:
| Persistent memory offers the cost and capacity advantages of storage devices at memory speed. It also brings opportunities for architecting software applications to utilize its new capabilities. Of course, the tradeoffs involved in using it are far from simple. NAND flash and other types of non-volatile memory behave quite differently from standard DRAM. Handling the differences requires new methods for system integration, programming, and management. Several solutions to the problems involved are currently in use, and several groups are proposing standards for both software and hardware. The obvious advantages of persistent memory indicate that designers will need to be aware of developments in this promising technology area. This seminar will provide an in-depth look into several leading technologies, current implementations, and applications that take advantage of persistent memory in compute and storage systems. Session PM-1: Introduction to Persistent Memory 8:30 -10:15 am Session PM-2 Current Implementations 10:30 - 12:00 noon
| | About the Organizer/Moderator: | |
|
, NaNth
1:00-5:00 PM | | Pre-Conference Seminar E: NVMe/TCP (Pre-Conference Seminars Track) | | Speaker(s):
| | Session Description:
| NVMe over Fabrics has quickly become a popular way to network flash memory, particularly with transports such as Fibre Channel, InfiniBand, and RoCE (RDMA over Converged Ethernet). However, the most popular networking fabric by far is the TCP used by standard Ethernet. Now NVMe/TCP provides access to this fabric, allowing data centers to combine what is now the standard SSD connection with the most popular network fabric. This is a winning combination because it is widely available and understood, offers very high bandwidth, and is inexpensive and easy to implement. Thanks to an optimized protocol stack, an end-to-end NVMe/TCP solution will reduce access latency and improve performance. Applications can then access SSDs rapidly, regardless of whether they are attached locally or accessed remotely across enterprise or data center networks. In the first half of this session, a tutorial covers NVMe over Fabrics, Ethernet with TCP, and NVMe over TCP. The second half is a hands-on tour of NVMe/TCP, the newest addition to the NVMe-oF family of standards and the first production-grade NVME/TCP-based software defined disaggregated storage solution, Lightbits LightOS. NVMe/TCP brings the power and speed of NVMe to TCP/IP networks and is already included in major Linux distributions. The audience will learn how NVMe/TCP works and how to connect to remote NVMe/TCP block devices and use them with applications. They will learn not just the theory of NVMe/TCP but also the practice of using it in a cloud data center. This is a hands-on session and attendees are expected to bring their own laptops. We will be using the Packet cloud environment for hands-on experimentation, bringing up Linux virtual machines that will use NVMe/TCP to connect to Lightbits LightOS servers in the Packet cloud.
| | About the Organizer/Moderator: | |
|
, NaNth
1:00-5:00 PM | | Pre-Conference Seminar F: Flash Storage Networking (Pre-Conference Seminars Track) | | Speaker(s):
| | Session Description:
| Large storage systems require more complex connections than simple interfaces to achieve maximum efficiency and scalability. Designers want to share storage readily among multiple compute nodes and be able to perform clustering, failover, and other system-wide operations. They thus need networked storage rather than the traditional direct-attached variety. However, networked storage introduces new tradeoffs in terms of cost, complexity, speed, latency, and other factors. Flash storage makes design even more difficult because of its higher bandwidth and performance, thus placing more strain on switches and adapters. Many connection technologies exist, including Ethernet, Fibre Channel, PCI Express, and InfiniBand. Higher-speed versions are available, as well as lower-latency extensions such as Remote Direct Memory Access (RDMA) and iSCSI extensions for RDMA (iSER). Typical issues governing selection include familiarity with particular approaches, existing use, latency and performance requirements, scalability, ecosystem (including management tools), and cost.
| | About the Organizer/Moderator: | |
|
, NaNth
1:00-5:00 PM | | Pre-Conference Seminar G: Designing Flash Controllers (Pre-Conference Seminars Track) | | Speaker(s):
| | Session Description:
| Flash controllers are complex systems that must handle a wide variety of effects and situations in solid state disks. They must implement properly the flash translation layer which converts system-level commands to instructions suited to the actual flash configuration. They must cope with wear, endurance, write amplification, and other issues. And they must be able to work with a variety of flash types and sizes to avoid getting into repeated design cycles. Error-correction techniques are still another issue with complex codes often being used to deal with failures in the underlying medium. FPGAs provide an obvious way to design initial versions and deal with low-volume situations. ASICs or market-standard chips can increase performance or reduce costs in many situations. Designs must also allow for later upgrades to lengthen their lifetimes.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-7:00 PM | | FMS Persistent Memory Programming Hackathon (Persistent Memory Track) | | Paper Presenters:
| | Session Description:
| The FMS Persistent Memory Programming Hackathon will provide a better understanding of how of how to use existing APIs to program persistent memory, as well as where it might benefit further research and development. Open Persistent Memory Hackathon drop-in sessions will run Tuesday August 6 and Wednesday August 7 from 8:00 am – 7:00 pm in the Great America Ballroom Foyer. Attendees will develop sample code based on open-source PM found in the Linux Kernel, PMDK.io and other interfaces. Following the Wednesday open Hackathon session, all are invited to a Persistent Memory Meetup from 7:00 – 8:30 pm in Great America Ballroom A to demonstrate the code they have developed.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-9:35 AM | | EMBD-101-A-1: Embedded Applications, Part 1 - Drive Design (Embedded Applications Track) | | Paper Presenters:
| Enhancing TLC Flash Designs for Embedded Applications
Thomas McCormick, Chief Engineer/Technologist, SwissbitImpact of the Usage models on the Storage devices definition
Ivan Ivanov, Distinguished Engineer, CoC Systems, HarmanBGA SSDs Lead the Way in Embedded Applications
Chris Lien, Senior Manager, ATP ElectronicsDesigning Reliable Flash Storage Systems
Axel Mehnert, Marketing Coordinator, Hyperstone | Session Description:
| Flash storage is becoming important factor for both consumers and in the enterprises. Benefits of flash storage are many, but much higher cost of flash storage as compared to traditional storage is making it hard to justify financially buying of the flash storage. What many storage customers don?t necessarily realized that price per megabyte or gigabyte of capacity is outdated model and it does not represent true cost of storage. Many enterprise consumers of storage keep buying more and more hard drives, not to satisfy capacity requirement, but to satisfy performance requirement, but when total cost of ownership is taken into consideration, flash storage becomes very competitive with much smaller footprint, cooling and power requirements. There are also certain workloads, which can benefit tremendously from using flash storage and those will be discussed in this presentation.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-9:35 AM | | FTEC-101A-1: Annual Update on Flash Technology (Flash Technology Track) | | Paper Presenters:
| Annual Update on Flash Technology
, , | Session Description:
| Is NAND flash at a crossroads? Is the end of scaling near? How is 3D flash doing? Are the reported production problems real and how important are they? What?s the real story with 3D XPoint? In practice, NAND flash technology keeps advancing with manufacturers reporting successful products at ever-smaller dimensions. 3-D flash is in production, as the transition continues. Flash memory remains the dominant non-volatile memory technology near term, and the 3D developments mean it will retain that status for years to come.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-9:35 AM | | INVT-101A-1: Gen-Z: The Best Interface for Emerging Memory Technologies (Enterprise Storage Track) | | Paper Presenters:
| Gen-Z: The Best Interface for Emerging Memory Technologies
Valerie Padilla, Technology Strategist, Server CTO, Dell EMC | Session Description:
| Most processor-memory interfaces assume a fixed type of volatile memory (usually DRAM). Other variations (such as static memory or ROM) require special interface hardware to operate on the memory bus. Gen-Z is a new data access fabric that abstracts the memory media, breaking the processor-memory interlock and allowing for a variety of memory types and operations. Memory may be persistent (that is, act like very fast storage), and operations may occur in it (that is, memory-centric or in-memory computing). Multiple paths may allow for multiple channels or protect against faults. With Gen-Z, new memory types can provide a wide range of functions for advanced applications without any need to slow down the memory bus or change its basic architecture. Memory and processors can now follow separate and independent innovation paths, increasing the capabilities of computers to handle new applications involving big data, real-time analytics, AI/ML, and edge computing.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-10:50 AM | | ARCH-101-1: Open-Channel SSDs for Host-Based Optimization (Architectures Track) | | Paper Presenters:
| Use of Open-Channel SSDs in Chinese Datacenters
Wei Xu, Technical Support Manager, Shannon SystemsA Global FTL Architecture to Drive Multiple SSDs
Roy Shterman, Senior Solutions Architect, Lightbits LabsOpen-Channel SSDs for Host-Based Optimization
Sheng Qiu, Staff Engineer, AlibabaAn Advanced Error Recovery Scheme for Open-Channel SSDs
Jeff Yang, Principal Engineer, Silicon MotionThe Denali Open-Channel Standard: Its Impact and Its Future
Javier Gonzalez, Principal Software Engineer, Samsung Electronics | Session Description:
| There are currently several alternative approaches to the handling of detailed SSD operations such as the placement of data and the timing of garbage collection. The standard approach has been to embed the details inside the Flash Translation Layer (FTL). The driver then only has to perform straightforward input and output operations. However, hyperscale operators want to be able to control everything from the system software in the interests of obtaining more efficient operations and minimizing latency. I/O determinism and the open-channel SSD offer the higher flexibility at the cost of more complex system software. Both paradigms make sense in different markets with clouds and other hyperscale operators preferring greater control while smaller data centers prefer the simpler interface. This session will focus on Open-Channel SSDs, which are able to reduce write amplification and improve QoS and NAND utilization in both data centers and enterprise solutions. However, various proprietary specifications have created some chaos in the industry as vendors are forced to support the same feature set on different (and incompatible) specification versions. This session will include Denali, the first industry effort to generate a standard Open-Channel specification as supported by Microsoft and a wide range of vendors and users.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-10:50 AM | | AUTO-101-1: Tomorrow's Auto Safety/Security Requirements (Automotive Applications Track) | | Paper Presenters:
| The Storage/Management Challenges of the Autonomous Transportation Ecosystem
Alan Messer, Chief Strategy & Technology Officer, InnovationShiftRequirements for Fail-Safe Automotive Storage Solutions with Secure NOR Flash
Sandeep Krishnegowda, Marketing Director, Flash Business Unit, Cypress SemiconductorThe Role of Secure Flash Memory in Automotive Applications
Anthony Le, VP Marketing, MacronixAutonomous Vehicles: Direction, Growth, and Challenges
Andrew Wygle, Embedded Software Engineer specializing in data programming of Flash devices., Data I/OTomorrow's Data Storage Integrity and Safety for Autonomous Cars
Bernd Niedermeier, Head of Automotive business develop, Tuxera | Session Description:
| The primary motivating forces for getting humans from behind the steering wheel of tomorrow’s vehicle is to reduce, if not eliminate, the approximately 1.3 million deaths on the world’s highways/byways today; reduce the volume of vehicles on the road; and significantly reduce vehicles’ environmental impact . Fast, reliable in-vehicle and ecosystem communications/data storage will be a key component in meeting that goal. Data in tomorrow’s vehicles must be 100 percent reliable, available and instantly provided to the various decision-making systems in the new moving environment. How will we design, develop and use storage redundancy, “soft fail” solutions that can protect passengers and those around them as well securely handle the constant flow of accurate information and vehicle upgrades/enhancements while the vehicle is in motion.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-10:50 AM | | CHNA-101-1: Flash Growth & Opportunity in China (China Track) | | Paper Presenters:
| Flash and NVM Technology Enabling AIoT
Michael Wang, VP, Gigadevice SemiconductorFlash Evolution Demands Controller Innovations
David Wu, CEO, YeeStor MicroelectronicsBalanced and Optimized Block Storage in the Tencent Cloud
Allen Liu, Solution Architect, Tencent CloudEthernet SSD (NVMe over Ethernet)
Hailuan Liu, R&D director, Hangzhou Dianzi UniversityThe Future of RRAM: From Embedded Application to In Memory Computing and Beyond
Jianguo Yang, Associate Professor, Institute of Microelectronics of the Chinese Academy of Sciences | Session Description:
| China is today a major player in both using and developing flash products. Those working at organizations outside China need to understand what is happening currently at Chinese companies, universities, and other organizations. The Chinese market offers many potential opportunities for products, services, and collaboration. Interests range throughout the non-volatile memory area, including development of new technologies, enterprise storage, embedded applications, and hyperscale datacenters.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-10:50 AM | | ENAP-101-1: Enterprise Applications, Part 1 (Enterprise Applications Track) | | Paper Presenters:
| Accelerating NVMe Adoption With Enterprise Storage-as-a-Service
Marc Leavitt, Sr. Director, Product Marketing, Zadara StorageIBM Storage: A Lighthouse for the Data-Driven Multicloud World
Eric Herzog, CMO and VP of Global Channels, IBMMulti-Site Collaboration Across Flash and Cloud with a Global Namespace
David Flynn, CEO, HammerspaceTipping Point: The Next Revolution for Compute and Storage
Steve Sicola, Founder & CEO, StorTrek Consulting | Session Description:
| Flash memory has revolutionized storage system and computing architectures for many enterprise applications. Actual case studies from innovative storage companies, including descriptions of the problem, approach, and results, provide examples of practical situations. Applications will include SQL and NoSQL databases, OLTP, data warehousing, big data analytics, Hadoop/MapReduce, financial transactions, and in-memory computing. Customers will co-present with some speakers.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-10:50 AM | | PMEM-101-1: Persistent Memory Part 1: Advances in Persistent Memory (Persistent Memory Track) | | Session Sponsor: SNIA, JEDEC, & OpenFabrics Alliance (OFA) | | Paper Presenters:
| State of the Union-Persistent Memory Today and Tomorrow
Jim Pappas, Director, Technology Initiatives, IntelCurrent Status and Overview of Standards Related to Persistent Memory
Jonathan Hinkle, Distinguished Member of Technical Staff, Micron TechnologyDeploying Persistent Memory and Acceleration via Compute Express Link
Stephen Bates, VP and Chief Architect of Emerging Storage Systems, HuaweiAn Overview of New and Emerging Media
Dave Eggleston, Principal/Owner, Intuitive Cognition ConsultingFPGA-Attached Persistent Memory Accelerates Applications Cost-Effectively
, , STT-MRAM, a High Performance Complement to Flash Memory
Tom Andre, Director of MRAM Circuit Design, HFC Semiconductor | Session Description:
| Persistent memory offers fast, byte-addressable access to data that persists! No more waiting endlessly for accesses to load data into memory or doing task switches to find something that can make progress during the wait. Obviously, the advance should mean huge speedups for applications that can’t currently keep data in memory, such as database, big-data analytics, and hyper-converged platforms. It is already proving to be great for accelerating applications such as MSFT SQL, VDI, high-availability storage, and others. But what about the software? It is currently all written and optimized to still copy data from memory to storage. Architects, developers, end user adopters, and vendors of today’s persistent memory must understand what development tools, platforms, management tools, and applications are currently available or in-progress to harness this new development in their systems.
| | About the Organizer/Moderator: | |
|
, NaNth
9:45-10:50 AM | | BMKT-101B-1: Annual Update on Flash Arrays (Business/Marketing Track) | | Paper Presenters:
| | Session Description:
| Flash arrays contain multiple SSDs, replacing traditional hard drives. They may be either all-flash with no hard drives at all or hybrids with both types of drives. They are currently very popular in data centers as a packaged solution that provides fast access at a reasonable price. They can do everything a disk array can do (and much faster), although SSDs usually have lower capacities than hard drives.
| | About the Organizer/Moderator: | |
|
, NaNth
9:45-10:50 AM | | DPRO-101B-1: Data Recovery of SSDs (Data Protection Track) | | Paper Presenters:
| The Past and Current State of Mobile Device Data Recovery
Steve Hruska, Mobile/Flash Data Recovery Manager, OntrackErasure verification on flash memory-Is the data really gone?
Will DeLisi, NAND Flash Recovery Specialist, DriveSavers Data RecoveryJosey Santana,
Data Recovery Engineer,
DriveSavers Data Recovery
The Use Of Encryption For Sanitization: A Case Study
Robin England, Senior Research and Development Engineer, OntrackIntra-disk RAID for Extreme Data Recovery
Cristian Zambelli, Assistant Professsor, University of Ferrara (Italy) | Session Description:
| SSDs are very reliable data storage devices. But when something goes amiss, users need to know what considerations should be taken into account to access and protect your organization's sensitive data? How is inaccessible data recovered from SSDs? Are SSDs designed to make finding lost data easier? When should one engage a data recovery specialist? How do you ensure your sensitive data is securely erased and sanitized from an SSD at the end of its useful life? Data recovery experts, digital forensic specialists, and SSD manufacturers will discuss the technology and science behind data recovery and data sanitization from flash based storage. If you've ever experienced data loss, want to know how to effectively plan and be prepared in the event of data loss, or need to know who the experts are and what they do, this is a must-attend session!. Attendees will gain an in-depth understanding of the effort involved in restoring lost data from an SSD, and how to ensure sensitive data is protected and erased when an SSD is at the end of its useful life or needs to be repurposed for alternate uses.
| | About the Organizer/Moderator: | |
|
, NaNth
9:45-10:50 AM | | EMBD-101B-1: Embedded Applications, Part 2 - Examples (Embedded Applications Track) | | Paper Presenters:
| Smart Storage Design for Edge Computing in Industrial IoT
Chanson Lin, Founder/CEO, Embestor TechnologyThe Role of Smart SSDs in the Artificial Intelligence of Things (AIoT)
Color Cheng, Manager, InnodiskUsing SSD Performance Numbers Effectively
Jan Peter Berns, Managing Director, Hyperstone | Session Description:
| Embedded systems have long used flash memory to support highly diverse system features and functionality under a wide range of operating conditions. In recent years, cost pressures from consumer applications have reduced the performance characteristics of the most widely available flash memory devices and have strained their ability to meet the more stringent reliability requirements of embedded applications. Embedded flash must have special characteristics to meet industrial, mil/aero, and other applications? needs in areas such as temperature, pressure, operating lifetime, shock, vibration, EMI, RFI, and radiation exposure.
| | About the Organizer/Moderator: | |
|
, NaNth
9:45-10:50 AM | | ENST-101B-1: Annual Update on Enterprise Flash Storage (Enterprise Storage Track) | | Paper Presenters:
| | Session Description:
| Enterprise flash storage continues to advance from being a faster plug-in replacement for disk drives to being a storage layer on its own. The original SSD conception obviously allowed storage designers to use the mature hardware and software ecosystem available for HDDs, adding only the Flash Translation Layer to make the necessary adjustments. However, this approach still left SSDs with all the disadvantages of disk storage including its complexity and low speed. NVMe has now brought a standardized high-speed parallel interface to solid-state storage, allowing it to act as a high performance tier in all-flash arrays, and now allowing it to be readily networked through the new NVMe Over Fabrics (NVMe-oF) standard. The world of NVMe-oF expanded rapidly this year from ultra-low latency JBODs to high-performance, low latency arrays with a full set of data services. Even more improvement is on the way, as we begin to treat solid-state storage as memory rather than as a peculiar form of disk drive. This advance will bring great change as in-memory databases become both truly all in memory and as persistent as traditional DBMSes. Low latency RDMA networks, which are also used by NVMe-oF, will allow an application to use memory semantics to write into another system?s non-volatile memory address space. This session will examine the development of these technologies into products enterprises can deploy, and the applications such products can best address. We?ll also look into our murky crystal ball a little and predict what users can expect to see over the next 2 - 3 years.
| | About the Organizer/Moderator: | |
|
, NaNth
9:45-10:50 AM | | INVT-101B-1: Achieving Lowest-Latency Storage with NVMe (NVMe Track) | | Paper Presenters:
| Achieving Lowest-Latency Storage with NVMe
Josh Goldenhar, VP Product Marketing, Lightbits LabsMotti Beck,
Director Enterprise Market Development,
NVIDIA
| Session Description:
| Low-latency storage is critical for a wide variety of applications, including transaction processing, financial trading systems, and video transmission. NVMe provides extremely low latency (single-digit microseconds) when used locally. How can we extend this level across networked flash which may experience both network delays and congestion? The solution is to provide software that pools NVMe storage access across the network at local speeds and latencies. The software must be distributed and support a variety of fabrics and architectures. It must also offer negligible overhead and application-independent operation. The result is NVMe sharing that scales performance and capacity linearly - and provides the low latency that today's applications require.
| | About the Organizer/Moderator: | |
|
, NaNth
9:45-10:50 AM | | NVME-101B-1: Zoned Namespaces Overview/Endurance Group Management (NVMe Track) | | Session Sponsor: NVM Express | | Paper Presenters:
| Handling Sequential Write Workloads in Mixed NVMe/Hard Drive Systems
Dave Landsman, Director of Industry Standards, Western DigitalEndurance Group Management: Host control of SSD media organization
Mark Carlson, Principal Engineer, Industry Standards, KIOXIAPaul Suhler,
Principal Engineer - SSD Standards,
KIOXIA
| Session Description:
| The NVMe Zoned Namespaces (ZNS) interface, which is being developed in the NVM Express workgroup, will be introduced. By dividing an NVMe namespace into zones, which are required to be sequentially written, ZNS offers essential benefits to hyper-scale organizations, all-flash array vendors and large storage-system vendors wishing to take advantage of storage devices optimized for sequential write workloads. ZNS reduces device-side write amplification, over-provisioning and DRAM, while improving tail latency, throughput and drive capacity. Also, by bringing a zoned-block interface to NVMe SSDs, ZNS brings alignment with the ZAC/ZBC host model already being used by SMR HDDs, and enables a zoned-block storage layer to emerge across the SSD/HDD ecosystem. SSD customers can have different requirements for the organization of the media in a drive: one large pool of capacity, separate sub-drives with performance isolation (I/O determinism), or one large pool plus a small pool capable of higher-performance writes. By allowing the host to configure a drive’s media in the field, a single SSD model can satisfy very different use cases. NVMe Endurance Group Management provides a mechanism for media to be configured into Endurance Groups and NVM Sets. This presentation will explain various use cases and how the mechanism is used to configure not just SSD media but also storage array components.
| | About the Organizer/Moderator: | |
|
, NaNth
9:45-10:50 AM | | TEST-101B-1: The Straight Truth: How Today's Storage Performs on Real Workloads (Testing Track) | | Paper Presenters:
| The Straight Truth: How Today's Storage Devices Perform on Realistic Workloads
Dennis Martin, Sr. Analyst, Principled Technologies | Session Description:
| Everyone (designers, marketers, salespeople, executives, and customers) wants to know how flash storage really performs. And they want to know it from an independent source using real-world applications. Principled Technologies will report on vendor-neutral performance tests run on database and virtualization workloads typical of today's data centers. The tests cover systems from several manufacturers, using a variety of form factors and interfaces and including both block and file protocols. Attendees will get good estimates of what to expect in practice, since the tests are independent and focus on current applications and environments. This session will also discuss recent advances such as NVMe over Fabrics (NVMe-oF), 3D XPoint performance, and high-speed interfaces such as 100 GbE and 32Gb Fibre Channel. Vendors will learn how their products shape up and where they should put their efforts.
| | About the Organizer/Moderator: | |
|
, NaNth
3:40-4:45 PM | | AUTO-102A-1: Autonomous Vehicles - The Storage Challenges of Edge Computing (Automotive Applications Track) | | Paper Presenters:
| Cross Temperature of NAND Flash Storage in Automotive Applications
Crystal Chang, Senior Manager, ATP ElectronicsMemory Requirements for Edge Computing in IoT applications
Nabil Damouny, Chief Strategist, Autonomous EdgeProviding Next-Generation Autonomy with PCIe Fabrics and Shared Storage
Wesley Yung, Principal Applications Engineer, Microchip TechnologyNetwork Drives Are the Next (R)evolution of Automotive Storage
Noam Mizrahi, CTO, Marvell | Session Description:
| With transportation edge computing takes on a new meaning. Will data volumes of data have to be stored/used close to the decision point or will only a minimum of data be required while the majority of data be captured, processed, used discarded. How much storage will be needed in tomorrow’s vehicle and how much will be required in the ecosystem. Find out the design requirements – size, performance, security, privacy - for the next level of edge computing storage in a mobile environment. Or miss the bus and watch the taillights of your storage opportunities disappear over the hill.
| | About the Organizer/Moderator: | |
|
, NaNth
3:40-4:45 PM | | BMKT-102A-1: Know What the Flash Customer Wants and Needs (Business/Marketing Track) | | Panel Members:
| | Session Description:
| Tech Target will present the latest 2019 research including IT purchase trends/growth/emerging techs/drivers and how broader infrastructure changes impact flash spending across the board. Tech Target will also include some best practices for marketing and sales leaders on topic alignment, best performing content etc. Following presentation is a Panel Session.
| | About the Organizer/Moderator: | |
|
, NaNth
3:40-4:45 PM | |
, NaNth
3:40-4:45: PM | | CMOB-102A-1: 5G Broadband & the Onslaught of Local Data (Consumer/Mobile Applications Track) | | Panel Members:
| | Session Description:
| 5G is coming to town. Its wide broadband throughput is like connecting a gigantic mile wide pipe to every home, business, smartphone, and tablet. 5G requires large local stores and local processing with low latency (below 1 ms). Different stakeholders have different needs for additional local storage and processing as a result of 5G becoming widely available. Attention moves from clouds to local devices and from remote processing to local processing with an increasing emphasis on very low latency. 5G will support many new applications with AI being pushed out to the edge as well.
| | About the Organizer/Moderator: | |
|
, NaNth
3:40-4:45 PM | | CTRL-102A-1: Annual Update on Flash Controllers (Controllers Track) | | Paper Presenters:
| | Session Description:
| The development of high-speed, well-designed, and low-cost controllers has been a key factor leading to wider use of flash memory. Controllers must account for flash's special features, as well as providing error correction, wear management, and high availability. The controller market has also been in a state of flux with new entrants constantly appearing while old stalwarts are acquired or lose their technical edge. The big news in the last year has been the rapid emergence of NVMe controllers, which appear ready to dominate the market. There has also been interest in higher-speed and lower-latency controllers and in ones that can be managed more precisely (via I/O determinism) and networked efficiently.
| | About the Organizer/Moderator: | |
|
, NaNth
3:40-4:45 PM | | ENST-102A-1: Enterprise Storage Design (Enterprise Storage Track) | | Paper Presenters:
| Dramatically Increasing Filesystem Performance with Flash
John Kim, Director of Storage Marketing, NVIDIAAdvancing NVMe-oF Adoption Based on Practical Experience
VR Satish, Co-Founder/CTO, Pavilion Data SystemsUsing Storage Class Memory to Accelerate All Flash Storage - Lessons Learned
Stephen Daniel, Distinguished Technologist, HPE / SPC | Session Description:
| Enterprise storage today must adapt to new requirements in the data center. Although cloud computing, analytics, deep learning, and big data have received the most attention recently, server and desktop virtualization, data tiering, efficient architectures, and implementation of data policies still must be further addressed. As the number of applications and users keeps increasing, the need for rapid, predictable, and intelligent access to data becomes more important. Solid state storage is a key technology in meeting higher demand and providing faster access to data at reasonable cost. However, managers must understand which flash technology to adopt and how to make price/performance tradeoffs. New physical architectures, new technologies, and new approaches to long-existing issues such as caching and distributed systems must all be leveraged to optimize enterprise storage and its solutions.
| | About the Organizer/Moderator: | |
|
, NaNth
3:40-4:45 PM | | INVT-102A-1: Key-Value Store and NVMe Deliver Close to DRAM Performance (Enterprise Storage Track) | | Paper Presenters:
| Key-Value Store and NVMe Deliver Close to DRAM Performance
Nikolas Ioannou, Research Staff, IBM Zurich Research Lab | Session Description:
| Many applications require low-latency data store services, a requirement typically satisfied using key-value stores backed by DRAM. However, large amounts of DRAM are very expensive. Also, recently introduced storage devices built on novel NVM technologies offer far higher performance than conventional SSDs. uDepot is a new key-value store built bottom-up to deliver the performance of fast NVM block-based devices. It is crafted carefully to avoid inefficiencies, uses a two-level indexing structure that adjusts its footprint dynamically to match the inserted items, and employs a novel task-based I/O runtime system to maximize performance. As an embedded store, uDepot's performance nearly matches the raw performance of fast NVM devices for both throughput and latency, while being scalable across multiple devices and cores. Using the popular Memcache protocol, it can deliver performance to network clients that is very close to DRAM-based systems. It offers much higher caching capacities at dramatically reduced cost in dollars per gigabyte. A uDepot-based memcache service is currently available as an experimental service in the IBM cloud.
| | About the Organizer/Moderator: | |
|
, NaNth
3:40-4:45 PM | | NVME-102A-1: NVMe-MI Overview & New Developments/NVMe-oF What's New in 1.1 (NVMe Track) | | Session Sponsor: NVM Express | | Paper Presenters:
| The NVM Express Management Interface: Overview & New Developments
Austin Bolen, Sr Principal Engineer, DellMyron Loewen,
Platform Architect in NVM Solutions Group,
Intel
NVMe Over Fabrics: What's NEW in Version 1.1
David Black, Senior Distinguished Engineer, Dell EMCFred Knight,
,
NetApp
Sagi Grimberg,
,
Lightbits Labs
| Session Description:
| The NVM Express Management Interface (NVMe-MI) specification defines a standardized framework for managing NVMe storage devices and enclosures. The specification defines an architecture, command set, and out-of-band transport mapping used by NVMe hosts or management controllers to manage NVMe elements. This presentation provides an introduction to the NVMe-MI 1.1 specification and an overview of new items being standardized for NVMe-MI 1.2. NVMe over Fabrics 1.1 is the latest version of the NVMe over Fabrics standard that accompanies the NVMe 1.4 standard. This combination introduces important new fabrics functionality such as the NVMe-oF/TCP transport and ANA (Asymmetric Namespace Access) Multipathing. Come hear about all the new fabrics functionality and how it’s likely to be used from three key authors of these standards.
| | About the Organizer/Moderator: | |
|
, NaNth
3:40-4:45 PM | | SECU-102A-1: Plugging Holes in Storage Security (Security Track) | | Paper Presenters:
| Advanced Countermeasures: Integrating SSDs Into Cyber-Security Defense
Sebastien Jean, CTO, Phison ElectronicsPhysical Chip-ID based Encryption and Security in SSD Controller
Hiroshi Watanabe, Professor, National Chiao Tung UniversityThe End of Opal 2.0 and SED SSDs?
Shing Lee, R&D Senior Manager, ADATA Technology | Session Description:
| This session will explore issues and challenges in securing data on flash memory devices and systems. It will pose related issues and opportunities and will discuss how technologies and market trends are impacting deployment of such systems in large government and private organizations. Topics covered will include self-encrypting storage, standards, secure erasure and elimination, encryption, trusted storage, embedded and industrial computing applications, and security for flash drives, SSDs, and industrial devices.
| | About the Organizer/Moderator: | |
|
, NaNth
3:40-6:00 PM | | ARCH-102-1: NVMe Zoned Namespaces for Improved Performance and Scalability (Architectures Track) | | Paper Presenters:
| Benefits of ZNS in Datacenter Storage Systems
Woo Suk Chung, Storage Software Team Lead, SK hynixNVMe Zoned Namespaces in Practice
Matias Bjorling, Distinguished Engineer and Country Manager, R&D Engineering, Western DigitalFast Integration and Furious Performance with Zoned Flash Drives
Robert Lercari, V. P. Engineering, Radian Memory SystemsSPDK Flash Translation Layer Library for Zoned Namespace SSDs
Wojciech Malikowski, Software Engineer, Intel Technology Poland | Session Description:
| Zoned Namespace (ZNS) SSDs are proposed for inclusion in the next NVMe specification. These will enable scalable storage architectures with a better TCO by shifting the main FTL functions to the host. This standards-based approach will enable host-side data placement to coalesce data writes at the application level, rather than at the individual SSD device level. ZNS SSDs will have improved management of data reads in concert with write operations for improved storage bandwidth performance, reduced write amplification, and efficient garbage collection. This architectural approach to large-scale storage repositories has further TCO benefits by allowing for the dramatic reduction DRAM used in SSDs, along with a reduction in NAND overprovisioning.
| | About the Organizer/Moderator: | |
|
, NaNth
3:40-6:00 PM | | PMEM-102-1: Persistent Memory Part 2: Software and Applications (Persistent Memory Track) | | Session Sponsor: SNIA, JEDEC, & OpenFabrics Alliance (OFA) | | Paper Presenters:
| Impact on Application Development - NVM Programming Model in the Real World
Andy Rudoff, Enterprise Storage Architect, IntelPersistent Memory Applications and Technologies
Mark Webb, Analyst, MKW VenturesProviding Native Support for Byte-Addressable Persistent Memory in Golang
Pratap Subrahmanyam, Fellow, VMwareApplying Persistent Memory to Neuromorphic Computing
Darryl Koivisto, Chief Technology Officer, Mirabilis Design | Session Description:
| Persistent Memory is READY. Industry standards are established, products are shipping (with more to come), platform support is available and multiple operating systems have native PM support. This enables new capabilities for a broad array of applications. Databases can now work faster and rebuild with little downtime. Run more virtual machines on the same core footprint. Accelerate cloud computing by removing file system overhead with direct persistent load/store semantics. Learn more about the latest software and application advances that have been enabled by the adoption of Persistent Memory, and see what others are doing to accelerate their application performance.
| | About the Organizer/Moderator: | |
|
, NaNth
3:40-6:00 PM | | SSDS-102-1: Enterprise SSDs (SSDs Track) | | Paper Presenters:
| The New Storage Tier between High-Performance SSDs and HDDs
Kent Smith, Product Marketing, Micron TechnologyHow to Detect and Handle SD and SSD Media Failures
Thom Denholm, Technical Product Manager, DatalightOvercoming the Reductions in NAND Endurance Ratings
JB Baker, VP Marketing, ScaleFluxApplying Blockchain to SSD Supply Chain Traceability
Michelle Lam, , IBMUsing Universal Backplane Management (SFF-TA-1005) to enable Tri-Mode Backplanes
Jeff Plank, Technical Director, Microsemi a Microchip CompanyDesigning SSD Storage Systems for Low Latency Without Large Outliers
Sebastien Jean, CTO, Phison ElectronicsTime to Market New Flash Device to Production SSD
John Plasterer, Chief Architect, NETINT Technologies | Session Description:
| Enterprise SSDs are achieving higher capacities and higher performance and are therefore used in more applications, displacing HDDs in a wide variety of environments. Storage system designers must make new tradeoffs among performance, data retention and endurance for SSDs to account for their special characteristics. Hyperscale data centers also have special requirements that must be met, most likely with architectures specifically intended for these fast-growing environments.
| | About the Organizer/Moderator: | |
|
, NaNth
4:55-6:00 PM | | AUTO-102B-1: Autonomous Vehicles Transportation Ecosystem Challenges (Automotive Applications Track) | | Panel Members:
| | Session Description:
| With transportation edge computing takes on a new meaning. Will data volumes of data have to be stored/used close to the decision point or will only a minimum of data be required while the majority of data be captured, processed, used discarded. How much storage will be needed in tomorrow’s vehicle and how much will be required in the ecosystem. Find out the design requirements – size, performance, security, privacy - for the next level of edge computing storage in a mobile environment. Or miss the bus and watch the taillights of your storage opportunities disappear over the hill.
| | About the Organizer/Moderator: | |
|
, NaNth
4:55-6:00 PM | | BMKT-102B-1: Data Growth and a Future with Zettabytes (Business/Marketing Track) | | Panel Members:
| | Session Description:
| A session on how AI/Machine learning, IoT and analytics will impact the explosive growth of data and how that will impact storage capacity and performance. “IDC says that 22 Zettabytes (ZB) of digital storage will be shipped across all storage media types between 2018 and 2025” –how will this change the way we look at and mange storage from both a capacity and performance standpoint?
| | About the Organizer/Moderator: | |
|
, NaNth
4:55-6:00 PM | | CMOB-102B-1: Break the Storage Bottleneck for 5G and IoT (Consumer/Mobile Applications Track) | | Paper Presenters:
| | Session Description:
| The Internet-of-Things (IoT) is here today. New 5G Low Power and private 5G networks enable sensors to store scads of information and transfer it faster and more efficiently. The IoT world has billion of devices with sensors, large storage, and communications. The result is massively connected networks, all active interacting and creating an incredible network intelligence. The low-power devices store GBs of data and do local processing to reduce the strain on networks. The challenge is to integrate IoT and 5G to provide benefits to the enterprise. NB-IoT has been announced but there are also private IoT and Long Range Bluetooth focused on IoT as well. Storage plays a critical role in all of these networks.
| | About the Organizer/Moderator: | |
|
, NaNth
4:55-6:00 PM | | CMOB-102B-2: New Form Factor Makes SSDs Shine in Client Applications (Consumer/Mobile Applications Track) | | Session Sponsor: Toshiba | | Paper Presenters:
| A New Replaceable SSD Form Factor
John Geldman, Director of SSD Industry Standards, KIOXIAA Replaceable Connector
Kenta Minejima, , Japan Aviation ElectronicsCompatibility Testing Environment for NVMe SSDs
Miki Takahashi, , Granite River LabsStorage Form Factor Trends for Notebook PC
Takashi R Sugawara, , Lenovo | Session Description:
| Replacement of hard drives by SSDs in client PCs is continuing, and SSDs have become standard in new computers. The M.2 form factor has ushered in a new class of thin, light SSD designs, improving the mobile computing experience. Next generation mobile devices embracing 5G, hybrid tablet/phone, and ultra-thin computing are driving the need for SSDs to become lighter, smaller, removable, and more power-efficient. With flash memory becoming even more dense and cost effective, a new form factor is needed to keep pace with new demands in client computing. See why this innovation will be used by leading system makers and how it will come to market.
| | About the Organizer/Moderator: | |
|
, NaNth
4:55-6:00 PM | | ENST-102B-1: Application Acceleration (Enterprise Storage Track) | | Paper Presenters:
| Applications Demanding Accelerated Disaggregated Storage
Scott Schweitzer, Sr. Manager, Marcom DCG, XilinxReal World Application Performance for the Latest Storage Technologies
Eden Kim, CEO, Calypso SystemsModelling and Measuring Demanding Transactional Workloads Using SPC-1
Stephen Daniel, Distinguished Technologist, HPE / SPCA Data-Path-Blind Distributed Storage Cluster Manager
Ronen Hod, Software Developer, Excelero | Session Description:
| One of the main purposes for employing flash memory is to accelerate applications. Many of the most popular recent applications, such as business analytics and NoSQL databases, often run quite slowly, particularly as their datasets increase in size. Furthermore, high-speed SSDs may themselves require acceleration as their requirements are beyond the capabilities of many low-end processors.
| | About the Organizer/Moderator: | |
|
, NaNth
4:55-6:00 PM | | INVT-102B-1: Achieving Fault Tolerance for Storage Class Memory (Persistent Memory Track) | | Paper Presenters:
| Achieving Fault Tolerance for Storage Class Memory
Huynh Tu Dang, Technologist, Western Digital | Session Description:
| Storage Class Memory (SCM) allows designers to replace several tiers of the memory hierarchy with a single, cost-effective, uniform memory/storage. To make such large-scale deployments practical, however, designers must be able to deal with unavoidable SCM wearout and failures. A new approach can provide fault tolerance in SCM-based main memory. It treats memory as a distributed storage system and relies on replication, plus programmable interconnect providing fast consensus to keep the replicas consistent. Initial experiments demonstrate reasonable overhead over local memory access and show great promise as scalable, fault-tolerant main memory.
| | About the Organizer/Moderator: | |
|
, NaNth
4:55-6:00 PM | | NEWM-102B-1: Annual Update on Emerging Memory Technologies (New Memory Technologies Track) | | Paper Presenters:
| Product Lifecycle Status of Emerging Memories
Mark Webb, Analyst, MKW Ventures | Session Description:
| Emerging memories have been forever banished to tiny niche markets subject to their impossibly insurmountable DRAM and NAND overlords? Or have they? Each year, Flash Memory Summit offers predictions about how new memory technologies are near the horizon and why they will disrupt the memory business and system architectures. Will emerging memories perpetually be 3 years to production or is the dawning of something new finally here? This session will offer a practical walkthrough of what we've learned about emerging memories over the past decade, the barriers that threaten their emergence, and what technical, manufacturing, ecosystem, and market conditions need to exist for them to be truly successful.
| | About the Organizer/Moderator: | |
|
, NaNth
4:55-6:00 PM | | NVME-102B-1: NVMe Software Drivers: What’s New and What’s Supported? (NVMe Track) | | Session Sponsor: NVM Express | | Paper Presenters:
| NVMe Software Drivers – What's New and What's Supported?
Scott Lee, Software Engineer Lead, MicrosoftMurali Rajagopal,
Director of Technology,
VMware
Jim Harris,
,
Intel
| Session Description:
| NVMe Driver eco-system exists for all leading Operating Systems including Microsoft Windows, Leading Linux flavors, VMWare as well as Platform Firmware UEFI. This session will cover what is new since last year for each of the drivers as well as gives an overview of the NVMe Driver eco-system including NVMe over Fabrics.
| | About the Organizer/Moderator: | |
|
, NaNth
4:55-6:00 PM | | SECU-102B-1: Building Security into Your System (Security Track) | | Paper Presenters:
| Protecting the Platform through Measurement and Attestation
Jeff Plank, Technical Director, Microsemi a Microchip CompanyCombating Persistent Cyber Attacks: The Secure Flash-to-Cloud Approach
Naseem Aslam, Sr. Manager, Product Marketing, Cypress SemiconductorYoni Kahana,
VP Customers,
NanoLock Security
Moving Target Defenses for Data Storage Devices
Donald Matthews, President and CEO, NexiTech | Session Description:
| Ransomware is malicious software that threatens to publish the victim?s data or block access to it unless the victim pays a ransom. The problem of ransomware began around 2012 and has grown rapidly (and internationally). The sophistication of the attacks has grown rapidly, and the difficulty of stopping them or tracing what happens to payments has also increased greatly. The largest threat appears to be to medium-sized installations that depend on their data, lack much sophistication in methods or personnel, and are capable of paying large ransoms. Examples include school districts, hospitals, professional firms, and small/medium-size businesses. As flash memory becomes more widespread, it will surely become still another method for attack. As a new technology, it is often left unprotected and its vulnerabilities are neither well-known nor well-understood. As more data migrates to flash, the problem will only get worse. Flash manufacturers at all levels will need to understand this threat and what they can do to combat it.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-7:00 pm | | FMS Persistent Memory Programming Hackathon (Persistent Memory Track) | | Paper Presenters:
| | Session Description:
| The FMS Persistent Memory Programming Hackathon will provide a better understanding of how of how to use existing APIs to program persistent memory, as well as where it might benefit further research and development. Open Persistent Memory Hackathon drop-in sessions will run Tuesday August 6 and Wednesday August 7 from 8:00 am – 7:00 pm in the Great America Ballroom Foyer. Attendees will develop sample code based on open-source PM found in the Linux Kernel, PMDK.io and other interfaces. Following the Wednesday open Hackathon session, all are invited to a Persistent Memory Meetup from 7:00 – 8:30 pm in Great America Ballroom K to demonstrate the code they have developed.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-9:35 AM | |
, NaNth
8:30-9:35 AM | | INVT-201A-1: NVMe-oF: What Performance Can You Expect for Real Applications? (NVMe-oF Track) | | Paper Presenters:
| NVMe-oF: What Performance Can You Expect for Real Applications?
Andy Walls, IBM Fellow, CTO FlashSystem, IBM | Session Description:
| NVMe inside storage and servers is becoming ubiquitous. Gartner is projecting that 89% of servers and 52% of storage systems will have NVMe attach by 2022. The advantages over disk interfaces are obvious - higher bandwidth and lower latency. Now we are seeing NVMe extended to provide system-wide attachment via NVMe-oF and a variety of fabrics. Here again, initial tests show significant advantages over previous networking schemes. Response time is generally quite close to that of attached storage. However, simple tests don't tell the entire story. What are the benefits with real systems that have significant parallelism and many operations going on at the same time? Networks introduce new problems such as traffic issues and noisy neighbors. Fabrics also vary in their maturity and capabilities. Progress is continuing in both network speeds and traffic control methods. Designers can expect NVMe-oF plus NVMe devices to provide outstanding overall performance, particularly in large hyperscale deployments.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-9:35 AM | | MRES-201A-1: Market Research Panel (Market Research Track) | | Panel Members:
| | Session Description:
| The flash memory market has recently seen explosive growth and rising demand, attracting many startups and creating exciting acquisitions and IPOs. On the technology front, NAND flash products are achieving lower cost through higher densities and by cascading from SLC to MLC to TLC to 3-D. New products are being targeted for the enterprise ranging from embedded flash on servers to improve their I/O performance a thousand-fold to networked SSDs targeting lower costs. A new generation of smart controllers significantly increases performance using hierarchical flash caching, improves high availability, and enhances endurance through deploying ECC, wear-leveling, over- provisioning, and garbage collection techniques together with auto-configuration and workload aware auto-tiering placement. Such controllers help make SSDs suitable for mission critical enterprise applications. Attend this key session on the state of the flash Industry. It will cover market forecasts, shares, technology progress, competition, go-to-market pricing structures, and possible mergers and acquisitions. Ask questions of analysts to crystallize your understanding of markets and potentially competitive products. Areas of interest include: Forecast of flash growth in server and storage markets Market segments where flash will displace other technologies New configurations such as all-flash arrays and flash in DIMM (NVDIMM) Price trends and their impact on markets New interfaces such as NVMe and NVMe-oF Effects of software-defined storage and hyperconvergence Whether future computer systems will incorporate flash directly at all levels The role storage will play in the transition from on-premise data centers to clouds The status of emerging technologies such as MRAM, RRAM, and 3D XPoint? Acquisition candidates
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-9:35 AM | | NVME-201A-1: NVMe IDC Market Overview/Enterprise Servers and Storage (NVMe Track) | | Session Sponsor: NVM Express | | Paper Presenters:
| NVMe Markets: Rising at Breakneck Speed with Even Better Days Ahead
Eric Burgener, Research VP, IDC | Session Description:
| NVMe is rapidly become a leading interface for enterprise storage, replacing the traditional hard drive interfaces such as SAS and SATA. Markets are now in the billions of dollars and show no signs of peaking. Recent primary research covers deployment in a wide variety of production environments and explores what is driving the purchases, what the major workloads are, and what issues have arisen. It includes an overview of the current market and future trends. Full steam ahead is the order of the day, and customers and vendors can use the research to guide new products and purchases, estimate market sizes, and project revenue opportunities. With the increased availability and awareness of NVMe products, more enterprises are making the transition to NVMe. Join our panel discussion to learn how server and storage vendors are helping make this transition possible.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-10:50 AM | | CTRL-201-1: Controllers and Flash Technology, Part 1 - Hardware and Algorithms (Controllers Track) | | Paper Presenters:
| Maximizing Performance/cost of SSD composed of Memory-type and Storage-type SCMs
Reika Kinoshita, Student, Chuo University/University of Tokyo (Japan)Advanced data integrity assurance for QLC flash
Licheng Xue, Sr. Staff Engineer, Starblaze TechnologyEnabling Fairness and Enhancing Performance in Modern NVMe Solid State Drives
Saugata Ghose, Special Faculty Systems Scientist, Carnegie Mellon UniversityTowards data-driven NAND flash controller development
Roman Pletka, Research Staff Member, IBM ResearchError Handling Technologies for QLC-Based Storage
Wei Lin, System Architect, Phison ElectronicsSSD with Compression: Implementation, Interface and Use Cases
Erich Haratsch, Senior Director Architecture, Marvell | Session Description:
| This session provides details on improving the endurance, retention, and performance of 2D and 3D NAND flash devices. Important strategies and signal processing that controllers can employ to further improve key reliability measures are revealed. The forum also presents novel implementations for reducing power usage and solving problems caused by write amplification. Learn about new technology developments and have time for questions and answers with top industry experts.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-10:50 AM | | ENAP-201-1: Enterprise Applications, Part 2 (Enterprise Applications Track) | | Paper Presenters:
| NVMesh: A Cluster Application For NVMe Drives
Tom Leyden, VP Marketing, ExceleroPersistent Memory: Revolutionizing the Modern Database
Gurmeet Goindi, Group Product Manager, GoogleAnalytics @ Rack-Scale with NVMe-oF
Walter Hinton, CMO, Pavilion Data SystemsOptimize Enterprise Applications with Shared Accelerated All-Flash Storage
Suhas Nayak, Director, Products & Solutions Marketing, Pure Storage | Session Description:
| Flash memory has enabled new storage system and computing architectures which can handle many enterprise applications far more efficiently than is possible with hard drives. This session will feature actual case studies by innovative storage companies, including descriptions of the problem, approach, and results. Customers will co-present with some speakers.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-10:50 AM | | NEWM-201-1: 3D XPoint: Current Implementations and Future Trends (New Memory Technologies Track) | | Paper Presenters:
| 3D XPoint Technology and Market Update
Mark Webb, Analyst, MKW VenturesWhere is 3D XPoint Headed?
Jim Handy, General Director, Objective AnalysisApproaches that Combine Optane with QLC Offer Solutions Competitive with TLC
Kapil Karkra, Principal Engineer, IntelMichal Wysoczanski,
Software Architect,
Intel
Piotr Wysocki,
,
Intel
Memory-Converged Infrastructure Based on Persistent Memory
Charles Fan, Co-founder & CEO, MemVergeUsing an Emerging Memory Cache with a High-Speed Client SSD
Stanley Huang, Director, SSD Product Marketing, Silicon MotionStorage Engine for 3D XPoint and 3D QLC NAND SSDs
Jack Zhang, Cloud Storage Architect, Intel | Session Description:
| The speculation surrounding Intel/Micron 3D XPoint technology has been huge, ever since its introduction in 2015. What is the reality? Where is the technology today and what applications are already using it? What are the effects of the relatively high prices (five times that of flash) that have now been suggested? Where will the technology be in 2022 and what steps will occur along the way? Obviously, this is long-term speculation with no guarantees. However, a look at the far horizon can provide guidance as to what developers are thinking and what initial users and observers now feel is possible.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-10:50 AM | | NVME-201-1: PCIe/NVMe Issues (NVMe Track) | | Paper Presenters:
| Boosting QLC SSD Performance and Endurance for Data Centers
Orit Wasserman, Sr Principal Software Engineer, Red HatStandardization for a Key-Value Interface underway at NVM Express and SNIA
William Martin, SSD IO Standards Development, Samsung ElectronicsUsing a PCIe Analyzer with NVMe-Based Products
Isaac Livny, Applications Engineer, Teledyne LeCroyNVMe Gen-4 Thermal Management: Too Hot To Touch
Grace Chen, Project Manager, Phison Electronics | Session Description:
| PCIe SSDs offer higher performance than ones based on disk interfaces, since they utilize the high-speed (and widely supported) PCIe bus. They have quickly become popular in a wide variety of enterprise applications, particularly in implementations utilizing the new NVMe standard. Of course, all the usual design problems occur ranging from connectors through power management, power consumption, configurability, and hardware/software tradeoffs. But with over 100 million enterprise PCIe ports already shipped, this is an approach enterprises find to be both reasonably priced and easily implemented. It can work in both client and data center applications.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-10:50 AM | | PMEM-201-1 Persistent Memory Part 3: Remote Persistent Memory-Case for Use Cases (Persistent Memory Track) | | Session Sponsor: SNIA, JEDEC, & OpenFabrics Alliance (OFA) | | Paper Presenters:
| The Case for Use Cases
Paul Grun, , RetiredThe Impact of Persistent Memory on Interconnects and Fabrics
Kurtis Bowman, Director Server Architecture/Technology, Dell | Session Description:
| Grappling with any emerging new technology requires a disciplined approach to developing the new technology and understanding its application. Remote Persistent Memory is an exciting new technology, but its ultimate impact is not yet in focus. This section describes how a ‘top down’ approach is being applied to the development of Remote Persistent Memory as an emerging technology. To illustrate that top down approach, this section introduces the work underway in the industry to describe use cases for remote persistent memory, and how those use cases will drive fabric requirements.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-10:50 AM | | SOFT-201-1: Composable Infrastructure and Software-Defined Storage (Software Track) | | Paper Presenters:
| Software Composable SSD Architecture Design for Enterprise Applications
Leander Yu, Vice President, Silicon MotionStorage Management with Swordfish API for Open-Channel SSDs
Slawomir Putyrski, Principal Engineer, IntelAddressing the Latency Gap with Composable Architectures
Larrie Carr, VP Engineering, RambusChallenges of Distributed Erasure Coding
Daniel Herman Shmulyan, Data Service Team Leader, ExceleroDAOS: Scalable Software-Defined Storage for HPC/Big Data/AI Convergence
Jeff Olivier, Software Engineer, Intel | Session Description:
| This session will include an exploration of composable infrastructure architectures that are enabled by technologies like OpenCAPI, CCIX and Gen-Z. It will also explore how software-defined infrastructures operate independently of any hardware-specific dependencies, and are programmatically extensible. In today's enterprise, the explosion of data and data sovereignty laws require infrastructures to be resilient and location transparent. Software infrastructures also allow for significantly improving operational efficiency. This session will discuss what is driving the move towards software-defined infrastructures, and how data management capabilities can be leveraged.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-10:50 AM | | TEST-201-1: Testing/Performance Analysis (Testing Track) | | Paper Presenters:
| Optimizing SSD Performance Efficiently with Realistic Workloads
Bob Weisickle, CEO, OakGate TechnologyEnterprise SSD Failure Analysis and Debug
Andrei Konan, Sr. Staff Firmware Engineer, SK hynixUsing an Analytics Pipeline to Monitor NAND Health Data in the Field
Behzad Amiri, Senior Nand Engineer, Pure StorageAccelerating the Qualification of Enterprise SSDs
Leah Schoeb, Sr. Developer Relations Manager, AMDOptimal Flash Storage Solutions for Applications Requiring Frequent Writes
KuoHua Yuan, Director, ADATA TechnologyCharacterizing SSDs with the SMART Monitor
Shuhei Tanakamaru, Senior Principal Engineer, Marvell Technology IncMemory Fencing for Detection of DMA Memory Address Violations
Bob Weisickle, CEO, OakGate TechnologyPerformance Evaluation of All-Flash Ceph Storage with QLC SSDs
John Mazzie, Senior Solutions Engineer, Micron Technology | Session Description:
| Testing and performance analysis are an essential part of flash development and system evaluation. Conformance to standards, behavior under environmental stress, and the effects of varying workloads and system conditions must all be checked thoroughly. Performance analysis is also a key to determining how devices will behave in operation and in comparing devices during the evaluation stage. One problem of particular concern is wear, since the underlying memory elements are known to fail after a certain number of writes have been performed. Evaluations must be done with workloads that reflect actual client experience to ensure validity.
| | About the Organizer/Moderator: | |
|
, NaNth
9:45-10:50 AM | |
, NaNth
9:45-10:50 AM | | INVT-201B-1: Hybrid Cloud Flash Optimization for AI and Metadata Management (AI/Machine Learning Track) | | Paper Presenters:
| Hybrid Cloud Flash Optimization for AI and Metadata Management
David Flynn, CEO, Hammerspace | Session Description:
| Flash technologies are key to enabling data-intensive workloads deployed across scale-out hybrid multi-cloud architectures, but as we approach the end of Moore's law we are forced to confront some uncomfortable truths. Advancements in silicon won't be enough on their own to carry us forward into the next era and building bigger, better data silos simply will not scale for next-generation workloads. To achieve our ambitions for scale, performance, and data governance we need to rethink the traditional approach to data management; becoming significantly more efficient with how data is made available and where it gets consumed by shifting our focus to the metadata. Managing rich metadata services at scale with an IT-assist from intelligent AI-driven automation will deliver the on-demand agility we need to seamlessly span clouds; the fine granular control we need to maximize the utility of our resources; and the intuitive usability we need for data operators to become masters of their data.
| | About the Organizer/Moderator: | |
|
, NaNth
9:45-10:50 AM | | MRES-201B-1: NVMe Market Research Panel (Market Research Track) | | Panel Members:
| | Session Description:
| NVMe has emerged rapidly as a major market since its introduction just a few years ago. It is intended as a standard package to enable designers to handle storage over the popular and widely supported PCIe bus. The largest part of the market is for simple adapters that connect a computer to a storage device. The advantage over the disk interfaces is higher speed. More recent extension of NVMe have opened new markets for switches, software, management, and system-level storage devices. Markets also include all-flash arrays, storage appliances, NVMe over Fabric (NVMe-oF) adapters, storage software, and intellectual property. NVMe-oF is the extension of NVMe to handle multiple computers, including clusters, multiprocessors, and distributed systems. As of 2016, over 70 companies had announced products based on NVMe, and over 100 companies had joined the NVM Express Organization, the governing body for NVMe. High-level members (known as promoters) include such important companies as Cisco, Dell EMC, Facebook, Intel, Micron, Microsemi, Microsoft, NetApp, Oracle, Samsung, Seagate, Toshiba, and Western Digital.
| | About the Organizer/Moderator: | |
|
, NaNth
9:45-10:50 AM | | NVME-201B-1: Client NVMe SSDs/NVMe in the Virtualized World (NVMe Track) | | Session Sponsor: NVM Express | | Paper Presenters:
| Client NVMe SSD Objectives and Opportunities
Shivashenkar Muralishankar, , IntelLee Zaretsky,
Distinguished Engineer,
Dell Technologies
NVMe in the Virtualized World
Suds Jain, Product Manager, VMware | Session Description:
| NVMe technology is continuously evolving to meet the to meet the needs of consumers in the client space. During this panel, the experts will discuss the major industry trends in client storage, provide an overview of the various client classes of SSDs and review relevant form factors. They will discuss the future of NVMe architecture in the client space, including the impact of PCIe 4.0 technology, TLC versus QLC and the status of hard drives. The data-centric world is challenging the current status-quo in IT infrastructure industry. NVMe and its ecosystem hold an important part of data-centric future. Please join this session to learn all about NVMe and NVMe-oF™ in virtualized world, how Virtual and cloud deployments are leveraging NVMe as a de-facto interface to access varied storage services and use-cases and how NVMe is fueling the evolution towards disaggregated infrastructure.
| | About the Organizer/Moderator: | |
|
, NaNth
9:45-10:50 AM | | SSDS-201B-1: New Enterprise and Datacenter SSD Form Factors (EDSFF) (SSDs Track) | | Paper Presenters:
| EDSFF 1 year in: What have we learned
Anthony Constantine, Platform Architect, IntelWhy m.2 is Unsustainable (and what can we do about it)
Lee Prewitt, Principal Hardware Program Manager, MicrosoftEDSFF for NVMe Computational Storage Processors
Stephen Bates, VP and Chief Architect of Emerging Storage Systems, HuaweiAdventures in Form Factors
Ross Stenfort, Storage Engineer, META | Session Description:
| It has been a bit over a year since the EDSFF form factor standards were launched in SFF. As the industry has begun building to the specs, we have learned some things, and already begun innovating on the specifications. This session will cover what we've learned, in general, since the introduction. We'll focus in particular on the evolution of E1.S to enable storage and compute acceleration use cases in compute servers. And we'll discuss how E1.S and E1.L, together, are providing solutions to replace M.2 in Datacenter and Enterprise.
| | About the Organizer/Moderator: | |
|
, NaNth
3:20-4:25 PM | | AUTO-202A-1: Auto Infotainment Advances/Obstacles (Automotive Applications Track) | | Paper Presenters:
| Auto Infotainment and Connectivity in Autonomous Vehicles - Impacts on Storage
Greg Basich, Associate Director, Automotive Infotainment and Telematics, Technology InsightsSecure F-RAM for Automotive Event Data Recording in ADAS Applications
Doug Mitchell, Product Marketing Engineer MPS, Cypress SemiconductorInfotainment and Autonomous Vehicles - The challenges of storage
Michael Huonker, Architect Infotainment, DaimlerNVMe Mass Storage Integration on Emerging Automotive Systems
Matthias Beste, HW Design Engineer, Intel | Session Description:
| Auto Infotainment store or stream Advances - Entertain or Distract - The infotainment system of tomorrow’s vehicle will be the major control session for the vehicle and passengers that will have to both inform/entertain them but also enable the vehicle and individuals to quickly and accurately analyze the vehicle as well as stationary and moving objects around them. How will we deliver, store and use this constant stream of data, information, entertainment efficiently and effectively to a centralized location. How much high-speed storage capacity will be needed and how will redundancy be implemented to provide sufficient safeguards to the vehicle all of the time and in all types of conditions.
| | About the Organizer/Moderator: | |
|
, NaNth
3:20-4:25 PM | | BMKT-202A-1: CMO Panel: Flash Will Be Everywhere but the Customer Still Rules (Business/Marketing Track) | | Panel Members:
| | Session Description:
| Marketing enterprise flash storage is difficult. The technology is complex and fast-changing, and customers are often very confused and ill-informed. How does one develop a clear message that meets the customer’s needs and offers solid ROI? What role does marketing analytics play in gaining a competitive advantage? Has the marketing playbook really changed? What are the winning strategies with the shift to cloud, hyperconverged architecture, and software defined infrastructure? What magic do successful marketers have up their sleeve as they follow the golden rule of “Know Your Customer”?
| | About the Organizer/Moderator: | |
|
, NaNth
3:20-4:25 PM | | CTRL-202A-1: CXL - A Coherent Interface for Ultra-High-Speed Transfers (Controllers Track) | | Panel Members:
| | Session Description:
| CXL introduces a general high-speed coherent interface, allowing the ultra-high-speed connection of all kinds of new technologies including accelerators, coprocessors, memories, and other devices. It thus supports various types of persistent memory, coprocessors (including GPUs and AI chips), and accelerators (including FPGAs and more advanced devices). CXL’s advantages include: · Single link for I/O, cache, and memory (including coherency) · Physical compatibility with the widely used PCIe interface with greatly reduced latency and coherency required by memory devices. · Support from major server, storage, software, memory, processor, and networking equipment makers, as well as hyperscale customers · Simple transition from traditional processor-memory systems · Ability to meet the needs of big data and high-performance computing, as well as those of emerging applications such as AI, ML, VR/AR, robotics, 5G, cloud computing, and image and video processing.
| | About the Organizer/Moderator: | |
|
, NaNth
3:20-4:25 PM | | INVT-202A-1: Handling the Network Requirements of High-Speed NVMe SSDs (Hyperscale Applications Track) | | Paper Presenters:
| Handling the Network Requirements of High-Speed NVMe SSDs
Rob Davis, VP Storage Technology - Networking Business Unit, NVIDIAManoj Wadekar,
Storage Architect,
Facebook
| Session Description:
| Today's NVMe SSDs can provide data at rates up to 20 Gb/s, with higher numbers coming. Such rates are essential to meet the requirements of applications such as real-time analysis, AI/ML, IoT, and high-performance video and wireless. However, they put tremendous strain on network resources. How do designers match the network to today's needs and tomorrow's expectations at a reasonable cost? Solutions include higher-speed Ethernet connections, rack-scale architectures, composable infrastructure and compute storage disaggregation to improve scalability and meet disparate application requirements. New storage stacks are also critical to take full advantage of SSDs, persistent memory, and high-speed switches and routers. Of course, all these approaches come with their own costs and limitations. Designers will need a deep understanding of network behavior to achieve cost-effective, scalable solutions that extend even to the largest hyperscale datacenters.
| | About the Organizer/Moderator: | |
|
, NaNth
3:20-4:25 PM | | NEWM-202A-1: RRAM (New Memory Technologies Track) | | Paper Presenters:
| RRAM Comes of Age
Shane Hollmer, VP Engineering, Adesto TechnologiesUpdate XP ReRAM Technology
Amigo Tsutsui, Senior Producer, Sony Semiconductor SolutionsMaking SiOx ReRAM a Cost-effective Embedded Memory
Amir Regev, CTO, Weebit-nanoTaOx-based ReRAM for Variability-Aware Approximate Computing
Chihiro Matsui, Assistant Professor, Chuo University/University of Tokyo (Japan) | Session Description:
| RRAM (resistive RAM) development keeps progressing, with companies using a variety of approaches to pursue potential applications in high capacity storage, mid-range nonvolatile caches, and low cost embedded solutions. Come hear industry leaders offer views on the current state of RRAM technology, the target applications, and the road to commercialization.
| | About the Organizer/Moderator: | |
|
, NaNth
3:20-4:25 PM | |
, NaNth
3:20-4:25 PM | |
, NaNth
3:20-5:45 PM | | CTRL-202-1: Controllers and Flash Technology, Part 2 - Error Correcting Codes (Controllers Track) | | Paper Presenters:
| Improving Quality of Service for 3D NAND SSDs Using LDPC Correction
Rino Micheloni, Fellow, Microchip TechnologyIntelligent Read Threshold Tracking to Mitigate Read Retry Requests
Oliver Hambrey, Research Engineer, SigleadUltra MMI: an LDPC decoder that doubles throughput at end-of-life.
Shiuan Hao Kuo, Supervisor Engineer, Silicon MotionA Low-Cost LDPC Solution for Flash Memory
Osso Vahabzadeh, Staff Design Engineer, Symbyon SystemsUsing Machine Learning Techniques to Reduce SSD Costs
Cloud Zeng, Technical Supervisor, Phison ElectronicsImproving Waterfall Performance of low-cost FAID LDPC Decoders
David Declercq, CTO, CodeLucida | Session Description:
| This session describes recent developments in implementing LDPC (low density parity check) and other codes for error correction in NAND flash memories. You will learn practical aspects of power-efficient LDPC implementations, handling new NAND technologies, and developing an erasure recovery scheme. Get important insight from industry experts and have time for direct questions and answers!
| | About the Organizer/Moderator: | |
|
, NaNth
3:20-5:45 PM | | FTEC-202-1: 3-D Flash (Flash Technology Track) | | Paper Presenters:
| Comparison of Current 3D NAND Chip and Cell Architectures
Jeongdong Choe, Senior Technical Fellow, TechInsightsComponent-Level Characterization of 3D TLC, QLC, and Fast SLC NAND
Patrick Breen, Flash Characterization Engineer, IBMCharacterization of 3D NAND Flash Memories beyond 1GT/s
Rino Micheloni, Fellow, Microchip TechnologyModeling and Mitigating Early Retention Loss and Process Variation in 3D Flash
Sauguta Ghose, Special Faculty Systems Scientist, Carnegie Mellon UniversityOnur Mutlu,
Professor,
ETH Zurich and Carnegie Mellon University
Measuring the Difficulty of Programming 3D NAND
Vic Ye, Manager, YeeStor MicroelectronicsA Deep Dive into 3D NAND Silicon
Jung Yoon, Distinguished Engineer and CTO, Supply Chain, IBM | Session Description:
| 3D NAND flash continues to advance and will soon dominate flash technology because of its lower cost and higher density. Issues in its widespread use include error-correction methods, manufacturing challenges, reliability, and lifespan. Manufacturers will need to continue to improve 3D processes to meet the needs of many applications. Situations requiring high reliability and extended lifetimes will require a great deal more development effort.
| | About the Organizer/Moderator: | |
|
, NaNth
3:20-5:45 PM | | NVME-202-1: PCIe/NVMe Storage (NVMe Track) | | Paper Presenters:
| Building NVMe storage systems with PCIe 4.0 embedded switch IP
Sampath Banka, Senior Applications Engineer, PLDABuilding Storage Systems for the Burst Blockchain Platform
Olga Buchonina, Founder, ActionSpotDaniel Jones,
,
Burst Software Developer
Getting the Most Out of QLC-based NVMe Storage
Andy Watson, Startup Strategy, AWSFTL Flow Control for CFexpress Camera Hosts Using Large Sequential NVMe Reads
Vishwas Saxena, Technologist, Firmware Engineering, Western DigitalApplying NVMe in Embedded and Mobile Applications
Horace Chen, Director, Product Management, Phison ElectronicsSebastien Jean,
CTO,
Phison Electronics
Thunderbolt 3: Bringing NVMe Storage to Consumer Computers
Shailendra Sinha, Sr. Business Development Manager, IntelPhysical Partitioning NVMe Controller Design for Improved SSD QOS Consistency
Gary Adams, Associate VP of Enterprise Marketing, SMI | Session Description:
| PCIe SSDs have rapidly emerged as the devices of choice in the enterprise because of their high speed, well-understood and widely used interface, and extensive support from major vendors. The NVMe standard for storage operations over PCIe offers a base platform comparable to those available for disk interfaces such as SAS and SATA. Furthermore, continuing advances in the underlying PCI Express interface (now in Version 4.0) offer a solid path for the future. Attention has now moved to implementing a wide range of essential system-level features in PCIe/NVMe. These include server storage platforms, standards for remote monitoring and management, high-availability features such as dual-porting, and virtual implementations for use with VMware.
| | About the Organizer/Moderator: | |
|
, NaNth
3:20-5:45 PM | | PMEM-202-1: Persistent Memory Part 4: Current Research in Persistent Memory (Persistent Memory Track) | | Session Sponsor: SNIA, JEDEC, & OpenFabrics Alliance (OFA) | | Paper Presenters:
| Exploiting Managed Language Semantics to Mitigate Wear-Out in Persistent Memory
Shoaib Akram, Assistant Professor, Australian National UniversityPMTest:-A Fast and Flexible Testing Framework for Persistent Memory Programs
Samira Khan, Assistant Professor, University of VirginiaPerformance Characterization of a DRAM-NV Hybrid Memory Architecture for HPC
Brad Settlemyer, R&D Scientist, Los Alamos National LaboratoryHyperdimensional Computing & Cognitive Memory of the Pattern Kind
Gil Russell, CTO, WebFeet Research | Session Description:
| Research on persistent memory has exploded over the last five years. Top universities, industrial research labs, and device manufacturers have scholars and staff dedicated to subjects like advances in memory devices or memory cell design, operating system and file system designs for non-volatile memories, and implications of non-volatile memories for scientific, big data, and high performance workloads. This session presents some of the most recent research by leaders in the field.
| | About the Organizer/Moderator: | |
|
, NaNth
3:20-5:45 PM | | SOFT-202-1: Key-Value Store and Linux Technologies (Software Track) | | Paper Presenters:
| Approaching 10M I/Ops on a Single CPU Core
Ben Walker, Storage Solutions Architect, IntelDistributed Key-Value Stores: Performance and Scalability for Flash Media
Richard Elling, Principal Arcihtect, Viking Enterprise SolutionsSmart Key-Value Storage for 3D XPoint and 3D QLC SSDs
Jack Zhang, Cloud Storage Architect, IntelKey-Value SSDs: Design Overview and Use Cases
Stanley Miao, Chief Engineer, Shannon SystemsImproved Flash Performance Using the New Linux Kernel I/O Interface
Vishal Verma, Performance Engineer, IntelThe Coming Revolution to Deploy Storage Processors for Application Acceleration
Mark Mokryn, VP Products, Pliops | Session Description:
| Key-value (K-V) store is a data storage technique that can be as simple as a hash table, and which is often a distributed storage technique. A popular K-V store system is RocksDB, which was developed by Facebook. The LSM-Tree (Log-Structured Merge Tree) technique is used by K-V databases such as RocksDB, MongoDB and Cassandra. Clusters rely on distributed K-V stores for service and configuration discovery, and for sharing consistent state across the members and clients and for distributed locking. This session will explore issues including the impact of network performance as well as different kinds of memory used to support Persistent Memory in order to facilitate effective K-V databases.
| | About the Organizer/Moderator: | |
|
, NaNth
4:35-5:40 PM | | AUTO-202B-1: Auto Roundtable (Automotive Applications Track) | | Panel Members:
| | Session Description:
| The Storage/management challenges of the autonomous transportation ecosystem – What types of data do vehicles need when they are in motion from the physical environment – streets, roads, signals, signs – and in what form? How much of the data needs to be saved by the vehicle to “learn” so it can anticipate wants/needs of the people/products being transferred? Will most of the data be held in the cloud and used/discarded while the vehicle is in motion or should it be retained in the vehicle? Who will have access to that data and who will be responsible for keeping it safe, secure, private?
| | About the Organizer/Moderator: | |
|
, NaNth
4:35-5:40 PM | | CTRL-202B-1: Gen-Z Interface (Controllers Track) | | Paper Presenters:
| Gen-Z: Built-in Security for the Data-Centric World
Michael Krause, Lead Architect, Gen-Z ConsortiumIntegrating the Gen-Z Interface into Your Applications
Bastien Heneffe, , PLDA | Session Description:
| Gen-Z is a new data-access technology that provides high-speed, low-latency access to data and devices via direct-attached, switched, or fabric topologies. It utilizes memory-semantic communications to move data between memories on different components with minimal overhead. Gen-Z components use low-latency read and write operations to access data directly, and can move data with minimal application or processor involvement. Gen-Z delivers maximum performance in a modular architecture without sacrificing flexibility, and offers built-in, component-level security.
| | About the Organizer/Moderator: | |
|
, NaNth
4:35-5:40 PM | | FTEC-202B-1: Annual Update on Flash Memory for Non-Technologists (Business/Marketing Track) | | Paper Presenters:
| Annual Update on Flash Memory for Non-Technologists
George Crump, CMO, StorONE | Session Description:
| "Why a four-year-old child could understand this report. Run out and find me a four-year-old child. I can't make head nor tail out of it."- Groucho Marx in Duck Soup Do you feel like the storage industry in which you participate every day is passing you by? Are you not sure about the meaning of QLC, NVDIMM, RoCE, and SDS ? and afraid to ask? Do you think I/O determinism might have something to do with Darwin? If so, come let an expert speaker tell you what you really need to know about the very latest flash technologies. You?ll then be ready for the next round of revolutionary, disruptive, transformative, paradigm-shifting, and singularity-destroying advances that will come our way next year!
| | About the Organizer/Moderator: | |
|
, NaNth
4:35-5:40 PM | | ENST-202B-1: Flash in Cloud Computing (Enterprise Storage Track) | | Paper Presenters:
| The Future of Data-Center Infrastructure
Rajan Goyal, CTO, FungibleIBM Storage: A Lighthouse for the Data-Driven Multicloud World
Eric Herzog, CMO and VP of Global Channels, IBM2019: The Year of NVMe
Narayan Venkat, VP, Data Center Systems, Western DigitalOpen Composable Disaggregated Infrastructures for the Edge
Mark Miquelon, Partner Alliance Director, Western DigitalExpanding Non-volatile Memory In Heterogeneous Systems Using the CCIX Standard
Ravi Kiran Gummaluri, , CCIX/Xilinx | Session Description:
| Storage is a crucial part of any cloud, regardless of its type or ownership. Cloud service providers have become major players in buying flash storage technology, looking for more IOPS even at higher cost to provide better performance. Providers also advertise SSDs as being available to users at extra cost Cloud providers also like the higher reliability of flash memory and its operating efficiency (space, power, and cooling requirements). However, the extra cost remains an issue, particularly in the amounts needed in cloud environments, as do the issues of wear and endurance.
| | About the Organizer/Moderator: | |
|
, NaNth
4:35-5:40 PM | | INVT-202B-1: Flash Solutions for the Skyrocketing AI/ML Applications Market (AI/Machine Learning Track) | | Paper Presenters:
| Flash Solutions for the Skyrocketing AI/ML Applications Market
Radoslav Danilak, CEO, Tachyum | Session Description:
| The massive and growing demand for AI/ML applications across all industries is generating requirements for a broad spectrum of storage solutions across multiple use cases. In this rapidly changing environment, storage designers must address an entirely new series of challenges and problems. AI/ML programs require huge data sets and present increasingly challenging bandwidth and latency needs. Training, inference, and computation have their own unique issues, including large numbers of small files, and the emerging requirement for at least an order of magnitude increase in I/O performance from storage systems. New AI chips, as well as GPUs and other high performance processors, will put tremendous pressure on such systems to fill their pipelines and keep them busy. New architectures, new interconnects, and new memory types will all be needed to capitalize fully on opportunities in this multi-billion dollar market. The industry must rise to the occasion with new approaches and capabilities to unleash the full power of AI in tomorrow's marketplace.
| | About the Organizer/Moderator: | |
|
, NaNth
4:35-5:40 PM | | NEWM-202B-1: MRAM (New Memory Technologies Track) | | Paper Presenters:
| Overcoming Challenges in MRAM High-Volume Manufacturing
Kevin Moraes, Vice President, Applied MaterialsSTT-MRAM: A High Density Persistent Memory Solution
Sanjeev Aggarwal, Vice President of Technology, EverspinEnabling MRAM for Applications
Danny Sabour, VP of Marketing, Avalanche TechnologyMRAM: Memory for the Edge – And Beyond!
Jeff Lewis, SVP Business Development, Spin Memory | Session Description:
| MRAM is rapidly entering the marketplace with the production of 64Mb standalone STT-MRAM, and the integration of embedded MRAM integrated with 28nm logic. The world?s fastest SSDs leveraging a MRAM cache have been announced, and non-volatile (NV) logic with MRAM elements has been proposed. The distinguished panelists will discuss the enormous range of MRAM applications, the current state of development and commercialization, and the disruption and opportunities MRAM creates.
| | About the Organizer/Moderator: | |
|
, NaNth
4:35-5:40 PM | | NVME-202B-1: Leveraging NVMe-oF for Existing and New Applications (NVMe Track) | | Session Sponsor: NVM Express | | Panel Members:
| | Session Description:
| In this panel session we will discuss application and use case examples leveraging NVMe-oF. Which improvements should you expect with NVMe-oF and how does this apply to different types of applications? Learn from early adopters’ implementations of NVMe-oF, and what you need to consider when planning to migrate existing applications from SCSI to NVMe or deploying new applications with NVMe-oF. We will complete the session with a peek into the future of how NVMe-oF can enable new application use cases.
| | About the Organizer/Moderator: | |
|
, NaNth
7:00-9:00 pm | | IEEE AI, ML, and Storage Seminar (AI/Machine Learning Track) | | Paper Presenters:
| Using RISC-V to Accelerate Machine Learning
Zvonimir Bandic, Sr Director of Next Generation Platform Technologies, Western DigitalUnderstanding Data at Storage Edge: An AI / ML Perspective
Ned Varnica, Director, MarvellAnalog Computing for AI/ML Using Embedded Flash
Seung-hwan Song, CTO and Co-Founder, ANAFLASH | Session Description:
| This session covers how the RISC-V revolution is enabling new accelerators for machine learning (ML), the importance of the computation engines that understand the stored data, the new implementations of these computation engines at the storage edge, and a different approach than the digital logic-based computation units for artificial intelligence (AI) accelerators. This approach allows flash memory-based analog computing applications to reduce the cost and power of AI and ML accelerators. AI in its current nascent form of Artificial Narrow Intelligence is the third era of computing. The ML and AI workloads are computationally very expensive, and need radical compute, memory and storage architectures to reduce cost and power dissipation, and to enable wide-scale deployment both in the cloud and at the edge. At the same time, these architectures need to be scalable and flexible to be relevant for emerging AI and ML workloads. This seminar was organized by IEEE and ValleyML.ai, and is hosted as an FMS free event/open session. It features three talks from industry leaders who are working at the intersection of ML and AI with storage.
| | About the Organizer/Moderator: | |
|
, NaNth
7:00-9:00 pm | | Persistent Memory Meetup (Persistent Memory Track) | | Paper Presenters:
| | Session Description:
| Join SNIA for a lively gathering of Persistent Memory enthusiasts to share results of the two day FMS Persistent Memory Hackathon, where software developers gathered in a live coding exercise to better understand the various tiers and modes of persistent memory and explore existing best practices. All are welcome!
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-9:35 AM | | COMP-301A-1: Computational Storage - Controllers & Technology (Computational Storage Track) | | Session Sponsor: SNIA | | Paper Presenters:
| Transforming Storage Controllers With Low-Cost, Low-Power Compute
Neil Werdmuller, Director of Storage Solutions, ArmJason Molgaard,
Principal Engineer,
AMD
Using Native NVMe-oF SSDs to Advance Computational Storage
Shahar Noy, Marketing, MarvellImplementing Computational Storage with Existing SSD Controller Resources
Ramyakanth Edupuganti, Staff Applications Engineer, Microchip Technology | Session Description:
| Computational Storage is a paradigm shift in data center infrastructure that addresses the market demand to process huge datasets with minimal latency. Computational Storage use cases range from database, big data, AI/ML and Edge applications where there are traditionally large gaps in compute vs. the amount of data that needs to be processed. The framework for Computational Storage is driven by SNIA and the NVM Express standards groups. You will hear from vendors who will discuss their current and future offerings and end results from real usage models with real customer discussions.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-9:35 AM | | FSOL-301A-1: Flash Solutions for Oracle (Flash Solutions Track) | | Panel Members:
| | Session Description:
| All enterprise storage is not the same. Oracle has unique performance characteristics along with customer requirements with SLAs that can achieve the highest levels of availability, capacity optimization and scalability. Find out the unique capabilities of these vendors and how they meet these requirements for enterprise database deployment.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-9:35 AM | | INVT-301A-1: Memory-Centric Computing in the Big Data Era (Enterprise Applications Track) | | Paper Presenters:
| Memory-Centric Computing in the Big Data Era
Onur Mutlu, Professor, ETH Zurich and Carnegie Mellon University | Session Description:
| Ever-increasing amounts of data create a major challenge for computing systems. How can data centers meet the current and future demands of real-time analysis, AI/ML, image and video processing, cloud computing, graph analytics, genomics, and other data-intensive applications at reasonable cost? A key problem is that simply moving data around, such as from storage/memory to a CPU or from local sources to a cloud, can take a very long time. Just try moving a petabyte to another computer or to backup! Or try copying a few huge pages within main memory! Promising ways to reduce the number and size of large data transfers include: 1) adding processing capabilities to storage/memory devices to handle common functions locally, and 2) using new memory technologies to accelerate data-intensive applications and operations. The resulting in-memory processing designs can provide high-bandwidth, low-latency, low-energy, and low-cost solutions across a wide range of applications.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-9:35 AM | | INDT-301A-1: CTO Panel (Industry Trends Track) | | Panel Members:
| | Session Description:
| CTOs and their staff do not have an easy life in the flash memory industry. Things are changing rapidly, and the road ahead is almost impossible to discern or understand. So what are these people thinking currently? What do they see as basic trends that will determine the course of nonvolatile memory technology? And what do they think are just transient issues that will soon be forgotten? What prized techniques can they recommend (such as crystal balls, Ouija boards, and tarot cards) for gauging the future?
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-9:35 AM | | MEDE-301A-1: Volumetric: True 3D Capture Is Upon Us (Media & Entertainment Track) | | Panel Members:
| | Session Description:
| Volumetric video, LiDAR, photogrammetry, and even light fields are storming onto the stage as they exit the R&D lab and move into production. This session will walk through the technologies, the strengths/weaknesses, and show examples of their use in real world productions. The technology produces creative/technical challenges many orders of magnitude beyond 4k vs. 8k and VR/volumetric creation can “quickly and easily” produce more than 10TB of content data that must be instantly captured, stored, used, saved.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-9:35 AM | | NEWM-301A-1: Life Beyond Flash - New Non-Volatile Memory Technologies (New Memory Technologies Track) | | Paper Presenters:
| Emerging Memories Find New High-Performance Applications
Jim Handy, General Director, Objective AnalysisTom Coughlin,
President,
Coughlin Associates
DDR5 NVRAM: A Non-Volatile DRAM Replacement Specification
Bill Gervasi, Principal Systems Architect, NanteroChallenges and Strategies for Mass Adoption of Emerging Non-Volatile Memories
Simone Bertolazzi, Engineer Analyst, YOLE GroupWhen It Comes to Emerging Embedded Memories, One-Size-Doesn't-Fit-All!!
Jack Guedj, CEO, NumemComputing In Memory with tri-gate SONOS Nonvolatile Multi-level Memory
Yasuhiro Taniguchi, CTO, Floadia | Session Description:
| Is there life after flash? Or will flash memory keep improving and dominate all NVM technology into the next decade? The panelists will peer into their crystal balls, and provide perspective on the great non-volatile beyond. They will provide insight and analysis on technology trends, disruption, singularities, product roadmaps and completion dates, and other memory issues that may go beyond human predictive capabilities. Bring your opinions, comments or Ouija board, tarot cards, fortune cookies, astrological instruments, tea leaves, or magic lamps and join in the discussion!
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-10:50 AM | | AIML-301-1:Using AI/ML for Flash Performance Scaling, Part 1 (AI/Machine Learning Track) | | Paper Presenters:
| A Scalable AI Data Pipeline for Storing and Processing Ingested Data
Sanhita Sarkar, Global Director, Western DigitalAccelerating NVM to GPU Transfer Rates Using PCIe Fabrics and Logical Volumes
Vincent Hache, Principal Applications Engineer, Microchip TechnologyManaging Massive Input Data in Flash for AI and Deep Learning Applications
Dejan Kocic, Sr Systems Engineer, NetAppReRAM for Implementing Neural Network Synapses
Amir Regev, CTO, Weebit-nanoExploring the Impact of System Storage on AI/ML Workloads via MLPerf
Wes Vaske, Principal Storage Solutions Enginee, Micron TechnologySearch Acceleration and Learning at the Edge with Crossbar ReRAM
Sylvain Dubois, , Consultant | Session Description:
| AI and machine learning (ML) are being used in many areas of the NVM industry. For example, various storage device health indicators can be monitored, and an ML model can be used to identify the drive's health by utilizing a deep neural network. However, such techniques require processing power and adequate storage apart from the storage that they are monitoring, and a Deep Learning (DL) approach requires the subsystem to learn by itself. This is just a sampling of the kind of issues addressed by the talks in this track.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-10:50 AM | | CTRL-301-1: Flash Controller Design Innovations (Controllers Track) | | Paper Presenters:
| Tunable Flash Translation Layer Improves Storage System Performance
Chris Bergman, Sr Firmware Architect, BurlywoodImproving the locality of generalized integrated interleaved codes
Xinmiao Zhang, Associate Professor, Ohio State University (ECE Department)A novel HW/FW co-designed SSD controller archtecture to boost up SSD performance
Wei Xu, Vice President, Maxio TechnologyRaw Access to Flash Memory Meets Special System Needs
Kenneth Pillay, CEO and Director, YALLIPTable-less Controllers for Persistent Memory in Datacenter Applications
Bernard Shung, President, Wolley | Session Description:
| NAND flash technology has produced continuously new records in the last years which have consolidated its position as the leading type of nonvolatile memory. Despite all advances, flash management demands increasing efforts from controllers. How can we unleash the full NAND flash bandwidth with minimal controller overhead? Can firmware designs keep up? Where do application and workload profiles meet controller design options? What controller breakthroughs are required for even faster non-volatile technologies such as MRAM, RRAM, 3D XPoint, and memristors? There are many possibilities out there, but the investment is high and the simple concepts have already been developed.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-10:50 AM | | NVME-301-1: PCIe/NVMe Technology Update (NVMe Track) | | Paper Presenters:
| The New Generation of Storage: PCI Express 4.0 and 5.0
Debendra Das Sharma, Chair, PCI-SIG Marketing Workgroup, IntelExtend the usage of Namespace: managing multi-media to improve app performance
Ron Yuan, R&D Vice President, MemblazeOffloading the NVMe SSD Datapath to Intelligent Controllers
Ziv Serlin, Principal Engineer, AmazonVMware's NVMe Support, and Forward Looking Plans
Murali Rajagopal, Director of Technology, VMwareThe Interconnect Dilemma: Next-gen Fabrics
Niraj Mathur, VP of Product, GigaIOScott Taylor,
Director Software Engineering,
GigaIO
Video Processing Secrets for NVMe
Daniel Zhou, , NVMe Use Case: Surveillance Video Processing
Tao Zhong, CTO, NETINT Technologies | Session Description:
| PCIe SSDs have rapidly emerged as the devices of choice in the enterprise because of their high speed, well-understood and widely used interface, and extensive support from major vendors. The NVMe standard for storage operations over PCIe offers a base platform comparable to those available for disk interfaces such as SAS and SATA. A further step in its acceptance is the development of a wide variety of added facilities. They include more efficient I/O frameworks (typically open-source), solutions for performance and power tradeoffs, ways to handle latency problems, and specially designed software such as file systems. Obviously, such added facilities grow the PCIe/NVMe ecosystem, further encouraging developers to use it in their storage systems.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-10:50 AM | | NVMF-301-1: Demystifying NVMe over Fabrics (NVMe-oF Track) | | Paper Presenters:
| Designing Fabric accelerators for NVMe storage arrays by using MRAM
Pankaj Bishnoi, Director of Applications, EverspinA Journey into NVMe-oF: Options, Trade-offs, and Challenges
Ihab Hamadi, Head of Systems Architecture, Western DigitalAchieving Scalable 5M+IOPS on NVMe/FC with Ultra Low Latency
Jayamohan Kallickal, Distinguished Engineer, BroadcomWenhua Liu,
Staff 2 Engineer,
VMware
End to End Gen4 NVMe-oF through PCIe Fabric and Smart NIC
Brian Pan, General Manager, H3 PlatformWhy cache in DRAM if you use NVMe-oF
Radu Stoica, Researcher, IBMNVMe-oF SAN telemetry with programmable switches
Surendra Anubolu, Distinguished Engineer, BroadcomFabrics/NVMe-oF's Storage's New Magic Wand
Arindam Sarkar, Director - Storage Engineering, MSys TechnologiesLeading Early Adopter Markets for NVMe-oF
Walter Hinton, CMO, Pavilion Data Systems | Session Description:
| The new NVMe SSD interface can be connected across a Fabric. In fact it can be connected across lots of different fabrics: Ethernet (3 approaches), Fibre Channel, InfiniBand, and PCIe to date. Data Centers want to share storage readily among multiple compute nodes and be able to perform clustering, failover, and other system-wide operations at NVMe SSD speeds. NVMe over Fabrics (NVMe-oF) is the solution. This talk will describe the technology in its many forms. Describe use cases, for both Enterprise and cloud, where it is being applied. Then finish with potential future directions it is heading.
| | About the Organizer/Moderator: | |
|
, NaNth
8:30-10:50 AM | | SOFT-301-1: Kubernetes and Cloud Software (Software Track) | | Paper Presenters:
| Leveraging Linux's Native Storage Functionalities via Open-Source
David Hay, Automation Engineer, Lithia MotorsAutonomous Kubernetes Flash Management with Multi-Cloud Application Awareness
Venkat Ramakrishnan, VP Engineering, PortworxRunning Stafeful Workloads on Flash and Kubernetes: Challenges and Solutions
Jagadish Mukku, Sr Architect, Robin.ioCo-Design Software and Hardware for SSD Storage
James Liu, Storage Architect, AlibabaData Orchestration for Kubernetes Stateful Tier-1 Primary Workloads
Marc Fleischmann, President & Founder, Datera | Session Description:
| Kubernetes is an open source system for management and automation of containerized applications. Container Storage Interface (CSI) provides a standard plug-in interface to expose block and file storage systems to container applications. OpenSDS (Linux Foundation project) provides a unified software-defined storage control plane for managing wide variety of storage backends such as Ceph, DRBD, NVMeoF targets. OpenSDS provides CSI based integration to expose flash block storage to Kubernetes container workloads. Running stateful workloads such as databases and big data on Kubernetes is hard. However, there is a universal need to complement its application-driven compute orchestration with complementary data orchestration. Storage stacks need to support more than the basic ability to provision storage volumes to Pods inside which stateful workloads run. Stateful workloads are performance-sensitive, requiring the correct binding between apps and flash storage to ensure predictable performance and higher utilization in a multi-tenant consolidated Kubernetes environment. The design of Kubernetes makes storage and data management very difficult for deploying stateful workloads such as Hadoop, Oracle, Splunk, Mongo, Elastic, SAP HANA, etc. This session will explore these and many other issues that are critical for supporting the Cloud.
| | About the Organizer/Moderator: | |
|
, NaNth
9:45-10:50 AM | | BMKT-301B-1: VC Forum (Business/Marketing Track) | | Panel Members:
| | Session Description:
| The flash memory area is bursting out all over with many startups and high-priced acquisitions. What are the short-term and long-term investment prospects? What are VCs looking for in funding flash storage companies and what do they think will be the key factors in achieving success? Is there enough room for all the current flash storage companies? Do the current and projected revenues justify the valuations? Which companies are most likely to succeed? When (if ever) will MRAM, RRAM, and other alternative NVM technologies be ready for prime time? What effect will the current industry storage industry slowdown have? How do such current hot topics as NVMe (and NVMe over fabrics), persistent memory, NVDIMM, and 3D XPoint? look as investment prospects?
| | About the Organizer/Moderator: | |
|
, NaNth
9:45-10:50 AM | |
, NaNth
9:45-10:50 AM | | ENAP-301B-1: How Flash Will Transform Enterprise Applications (Enterprise Applications Track) | | Panel Members:
| | Session Description:
| Flash memory has already increased the performance of enterprise applications tremendously. However, we have just begun to see the long-term effects. Flash will find even more uses with all-flash arrays dominating local storage, server-side and storage-side caches providing higher performance, and even cold storage moving to flash rather than disk or tape. Real-time analytics will be a key area of interest with both computational performance and access to big data being essential. New developments such as flash on the memory bus and persistent and storage-class memory will maker flash even more important. Data centers will need to take full advantage of flash memory at every level ? in computers, in servers, in networks, and in storage systems.
| | About the Organizer/Moderator: | |
|
, NaNth
9:45-10:50 AM | | FSOL-301B-1: Flash Solutions for SAP (Flash Solutions Track) | | Panel Members:
| | Session Description:
| Storage for SAP customer environments have unique requirements that go beyond just fast storage. Multi-tenancy, QoS, security and overall data protection must be considered as key building blocks of an SAP storage solution. Discover the unique and different approaches to these customer needs from innovative storage solution vendors.
| | About the Organizer/Moderator: | |
|
, NaNth
9:45-10:50 AM | | INVT-301B-1: Hot Topics in Academic Flash and NVM Research (New Memory Technologies Track) | | Paper Presenters:
| Hot Topics in Academic Flash and NVM Research
Saugata Ghose, Special Faculty Systems Scientist, Carnegie Mellon University | Session Description:
| Every year, hundreds of researchers in academia push the frontier of new ideas in the field of NAND flash memory, non-volatile memories, and storage architectures. This year, we are presenting a summary of the top works and trends in academia. A committee consisting of both industrial and academic experts have selected three standout works that were published in the last year, based on a combination of the future potential of the work and its research area, the novelty of the work, and the relevance of the work to industry. This talk will briefly discuss the selected works, put them in the context of larger projects and topics that are being worked on actively at universities across the world, and provide additional resources and contacts for anyone interested in learning more about the works.
| | About the Organizer/Moderator: | |
|
, NaNth
9:45-10:50 AM | | MEDE-301B-1: VR – Virtual Stories, Experiences You Control (Media & Entertainment Track) | | Paper Presenters:
| Immersive Media Filmmaking
Diego Prilusky, Founder/GM, Intel StudiosPractical Considerations of the Volumetric Filmmaking Pipeline
David Hoffman, Business Development Manager, Blackmagic DesignHigh Performance Computational Storage for Video, Volumetric and VR
John Plasterer, Chief Architect, NETINT Technologies | Session Description:
| While the excitement is clear for consumers and developers, we’re also seeing increased adoption of VR for commercial and business applications. The possibilities are endless, but first, we must solve one massive challenge currently being overlooked: What to do with the endless, enormous data files being created by VR platforms. A new video capture approach – volumetric – is poised to enhance and enrich 360- (spherical video) by delivering true realism where viewers actually move around in 3-D virtual spaces, interacting with 3-D video of real people. Think of the potential applications, the experiences, the volumes of data that will be captured/stored/used to make you the director.
| | About the Organizer/Moderator: | |
|
, NaNth
2:10-3:25 PM | | COMP-302A-1: Computational Storage: Implementation Methods (Computational Storage Track) | | Session Sponsor: SNIA | | Paper Presenters:
| Computational Storage: From Devices to Subsystems
Pankaj Mehra, VP Product Planning, Samsung ElectronicsAlibaba's In-Storage Computing Architecture and Ecosystem
Feng Zhu, Staff Engineer, Alibaba CloudEdge Computing with Advanced ARM based Data Center SOC and NVMe
Rick Carlson, CEO, AgylstorUse an Intelligent SSD to Accelerate Machine Learning
HungWei Tseng, Assistant Professor, University of California Riverside | Session Description:
| Computational Storage is a paradigm shift in data center infrastructure that addresses the market demand to process huge datasets with minimal latency. Computational Storage use cases range from database, big data, AI/ML and Edge applications where there are traditionally large gaps in compute vs. the amount of data that needs to be processed. The framework for Computational Storage is driven by SNIA and the NVM Express standards groups. You will hear from vendors who will discuss their current and future offerings and end results from real usage models with real customer discussions.
| | About the Organizer/Moderator: | |
|
, NaNth
2:10-3:25 PM | | ENST-302A-1: Designing Storage Systems for the Exabyte Era (Enterprise Storage Track) | | Paper Presenters:
| Designing Storage Systems for the Exabyte Era
Rob Peglar, President, Advanced Computing and Storage | Session Description:
| Storage has moved to center stage in the era of petabytes and exabytes. This little-studied area (who ever heard of a required course on it?) is now the center of attention as processors and networks have moved into a period of tremendous complexity and slow progress. To manage the huge rapid data flows required by key applications such as real-time analytics, AI/ML, virtual and augmented reality, exascale HPC, and IoT, significant system-level design breakthroughs are essential. Storage systems must take full advantage of all the latest technological advances such as persistent memory, high-density NAND (QLC), NVMe, and fabric-based networking. Scalability is obviously essential as datasets continue to grow at breathtaking rates, local stores and processing are necessary to avoid long data transfers, and flexibility is paramount as systems must take full control of both metadata and policies. System complexity must be reduced, particularly to keep the mega clouds manageable. International organizations at all levels must bring together researchers, designers, and users to create the standards, certifications, open-source software, and training required to make the new era of storage successful.
| | About the Organizer/Moderator: | |
|
, NaNth
2:10-3:25 PM | | FSOL-302A-1: Flash Solutions for Analytics (Flash Solutions Track) | | Panel Members:
| | Session Description:
| Data is the most valuable asset in modern enterprises, and data analytics is the key to unlocking critical insights to better service customer needs and to competitively differentiate products and services in the marketplace. Uncover the key selection criteria and storage solution considerations when evaluating all flash storage arrays.
| | About the Organizer/Moderator: | |
|
, NaNth
2:10-3:25 PM | | FTEC-302A-1: The Next Great Breakthrough in Non-Volatile Memory (Flash Technology Track) | | Panel Members:
| | Session Description:
| Advances in nonvolatile memory keep coming despite warnings about a slowdown or stoppage in technological innovation. Flash remains the dominant technology, and its run is likely to continue. Breakthroughs keep occurring, and advances raise density and speed and decrease costs. What will be next? Will it be multiple levels beyond QLC, a fourth dimension, smaller process dimensions, or something else? Of course, flash technology has been around for a while, and the next great breakthrough could be the long-awaited emergence of a major contender such as MRAM, RRAM, memristors, or carbon nanotubes.
| | About the Organizer/Moderator: | |
|
, NaNth
2:10-3:25 PM | | INDT-302A-1: Zettabyte Alert - Designing Storage for the Big Bang in Data (Industry Trends Track) | | Paper Presenters:
| | Session Description:
| IDC’s Digital Universe Study predicts that the amount of data generated will increase from 0.1 ZB in 2005 to 44 ZB in 2020 – and will grow further to 163 ZB by 2025. How will storage systems handle all that data while maintaining performance and keeping costs under control? What are the best ways for data centers to develop systems that are scalable yet feasible in an era of static budgets and headcount? What general guidelines must the storage designer consider? What are the key hints, warnings, tips, and tricks that will help product designers and data center managers alike prepare for the data onslaught?
| | About the Organizer/Moderator: | |
|
, NaNth
2:10-3:25 PM | | INVT-302A-1: Life After RAID: Big Data Requires a Better Approach (Enterprise Storage Track) | | Paper Presenters:
| Life After RAID: Big Data Requires a Better Approach
Mike McWhorter, Senior Technologist, Western Digital | Session Description:
| RAID has become a standard way of protecting data integrity. The various RAID levels provide many options for data centers to use with a wide range of cost/performance tradeoffs. However, traditional RAID is no longer capable of handling centers' needs now that data stores in the petabyte range are commonplace. Rebuild times have become way too long, and capacity requirements have become far too large. Furthermore, triple replication simply takes too much space to be economically feasible. The solution is erasure coding, which can survive more disk failures than RAID while using only a fraction of the space of triple replication. It allows users to keep rebuild times and storage costs under control for today's big data applications and for the zettabyte future.
| | About the Organizer/Moderator: | |
|
, NaNth
2:10-3:25 PM | | MEDE-302A-1: Living the Ready Player One Virtual Experience (Media & Entertainment Track) | | Panel Members:
| | Session Description:
| Find out what the entertainment and general business have in store for VR. Learn how the technology will prove to be a major asset in real estate, defense, healthcare, education and technology development. Architects, builders can enable clients to walk around and through projects even before ground has been broken. Military and aerospace organizations enable personnel to experience combat, equipment/systems and even venture through the space activities and worst-case scenarios safely. The enriched technology will help engineers to develop, test, work with, refine minute chips or entire systems in a virtual world they produce, refine without building/destroying a single solution. The technology will enable sporting enthusiast to be involved in all of the football, basketball, soccer action. Find out how our experts envision VR being used and explore how soon it can be part of your professional/personal life.
| | About the Organizer/Moderator: | |
|
, NaNth
2:10-3:25 PM | | NVMF-302A-1: Understanding NVMe over Fabrics on TCP (NVMe-oF Track) | | Paper Presenters:
| NVMe/TCP Interop
David Woolf, Senior Engineer, UNH-IOLAn NVMe/TCP Software-Defined Platform for Guaranteed QoS
Alex Shpiner, System Architect, Lightbits LabsAccelerating NVMe over TCP for Disaggregated Storage Applications
Tom Spencer, Senior Director Product Marketing, SolarflareUsing SmartNICs and Buffer Management to Improve NVMe over TCP Performance
Ron Renwick, Sr Director Product Mktg/SmartNICS, NVIDIAComparing NVMe-oF on RoCE vs. TCP
John Kim, Director of Storage Marketing, NVIDIA | Session Description:
| NVMe over Fabrics has been in the market over RDMA with Ethernet RoCE and InfiniBand since 2016 and Fibre Channel since last year. New to the market this year is NVMe over Fabrics on Ethernet TCP. This talk will describe the technology in its many forms. Describe use cases, for both Enterprise and cloud, where it is being applied. Then finish with Q/A to the four presenters.
| | About the Organizer/Moderator: | |
|
, NaNth
2:10-3:25 PM | |
, NaNth
2:10-5:00 PM | | AIML-302-1: Using AI/ML for Flash Performance Scaling, Part 2 (AI/Machine Learning Track) | | Paper Presenters:
| NVM Usage in the AI Era
Dave Eggleston, Principal/Owner, Intuitive Cognition ConsultingFlash Memory, Storage and Data Challenges for Production Machine Learning
Nisha Talagala, CEO, Pyxeda AIAn Analog Neuromorphic Compute System Based on SuperFlash
Mark Reiten, Vice President Licensing Division, Microchip TechnologyNon-Volatile Neural Network Accelerator in Your SoC
Sangsoo Lee, CEO, ANAFLASHComputation and Machine Learning in Flash Memory: Opportunities and Challenges
Jeremy Holleman, CTO, SyntiantPredicting I/O Patterns Using LSTM Neural Networks in Storage Systems
Ken Qing Yang, Chief Scientist, Dapu Microelectronics | Session Description:
| AI and machine learning (ML) are being used in many areas of the NVM industry. For example, various storage device health indicators can be monitored, and an ML model can be used to identify the drive's health by utilizing a deep neural network. However, such techniques require processing power and adequate storage apart from the storage that they are monitoring, and a Deep Learning (DL) approach requires the subsystem to learn by itself. This is just a sampling of the kind of issues addressed by the talks in this track.
| | About the Organizer/Moderator: | |
|
, NaNth
2:10-5:00 PM | | NVMF-302-1: Benefits and Use Cases for NVMe-oF (NVMe-oF Track) | | Paper Presenters:
| Deep Dive on Disaggregated Storage Architectures: Challenges and Solutions
Vaidyanathan Krishnamoorthy, Cloud Solutions Architect, IntelMrittika Ganguli,
Network Architect,
Intel
NVMe-oF: Benefits and Use Cases for Production Applications
Steve McQuerry, Sr. Technical Marketing Engineer, Pure StorageSmartNICs for E2E NVMe Disaggregation in a Software Defined Datacenter
Fazil Osman, Distinguished Engineer, BroadcomDeveloping NVMe-oF Composable Infrastructure with Virtual All-Flash Appliances
Alan Lin, Product Manager Lead, AccelStorImplementing Computational Storage in an NVMe-oF-Based System
Patrice Couvert, Technical Product Manger, KalrayUsing SmartNICs to build an NVMeoF Target Systrem
Ziv Serlin, Principal Engineer, AmazonWhat NVMe-Over-Fabrics Means to Software Defined Storage
Tom Lyon, Chief Scientist, DriveScaleBringing NVMe-oF to Hyperscalers with SmartNICs
Idan Burstein, Staff Silicon Architect, NVIDIA | Session Description:
| NVM Express over Fabrics (NVMe-oF) enables users to connect remote subsystems with a flash appliance to achieve faster application response times and better scalability across virtual data centers. NVMe over fabrics offers high transfer speeds, low latency, full standardization, and access to a large ecosystem. Data centers can employ it to get higher utilization, huge performance benefits, and high levels of scalability over hundreds or thousands or local and remote SSDs. It is thus well-suited to large sites such as clouds, megawebsites, and hyperconverged data centers.
| | About the Organizer/Moderator: | |
|
, NaNth
3:40-5:00 PM | | COMP-302B-1: Computational Storage: Applications (Computational Storage Track) | | Session Sponsor: SNIA | | Paper Presenters:
| Creating a Framework for Computational Storage Applications
Nick Adams, Principal Engineer, IntelDesign the Right Storage Systems to Accelerate Deep Learning
Jerome Gaysse, Senior Technology and Market Analyst, Silinnov ConsultingApplication System Design From Edge to Data Centers
Rohit Gupta, Segment Manager, Western DigitalComputational Storage Workloads - Implications for Datacenter Architectures
Jamon Bowen, Product Marketing Director, Xilinx | Session Description:
| Computational Storage is a paradigm shift in data center infrastructure that addresses the market demand to process huge datasets with minimal latency. Computational Storage use cases range from database, big data, AI/ML and Edge applications where there are traditionally large gaps in compute vs. the amount of data that needs to be processed. The framework for Computational Storage is driven by SNIA and the NVM Express standards groups. You will hear from vendors who will discuss their current and future offerings and end results from real usage models with real customer discussions.
| | About the Organizer/Moderator: | |
|
, NaNth
3:40-5:00 PM | | EMBD-302B-1: Flash Memory, IoT, and AI – Bringing It All Together (Embedded Applications Track) | | Panel Members:
| | Session Description:
| The Internet of Things (IoT) will generate far more data than people or even supercomputers can handle effectively with today’s methods. AI systems will make useful information out of continuous streams of IoT data. This capability comes with a significant increase in computational complexity and an increased demand on high-performance storage and processing power. The combination of flash memory, IoT, and AI is the way to produce business value out of this incredible new technology. And, of course, all this has to be done at reasonable cost and with reasonable power consumption. Case studies show the IoT strategy combining AI technologies that companies have developed to meet their business goals.
| | About the Organizer/Moderator: | |
|
, NaNth
3:40-5:00 PM | | FSOL-302B-1: Flash Solutions for AI/ML (Flash Solutions Track) | | Panel Members:
| | Session Description:
| Artificial Intelligence (AI) and Machine Learning (ML) are without a doubt the hottest things happening out there in the technology world. They are changing the way every industry is solving its biggest problems. Implementations stretch from mobile devices all the way to the data center with compute, memory, and storage needing to be balanced to achieve optimal performance. Uncover the key considerations when evaluating flash arrays to address these rapidly emerging applications.
| | About the Organizer/Moderator: | |
|
, NaNth
3:40-5:00 PM | | MRES-302B-1: Top 10 Things You Need to Know about Flash Memory Today (Market Research Track) | | Panel Members:
| | Session Description:
| Flash memory has morphed rapidly from an exciting new technology that had to be justified for specific use cases to a standard part of every data center. What other kind of storage would you be buying today? So what do vendors and users alike need to know about flash? Is it the emergence of 3D technology, the rapid rise of the high-speed NVMe standard, the promise of persistent memory (storage at memory speeds), the role of flash in scalable systems such as clouds and megawebsites, new methods for flash storage networking (such as NVMe-oF), ways to make software take advantage of flash memory, or large, hierarchical storage systems that cover everything from high-speed cache to long-term archiving? Our top industry experts will present a few of their own candidates for the Top Ten list. We will then open nominations to the audience and finish with our vote for the Top Ten for 2019.
| | About the Organizer/Moderator: | |
|
, NaNth
3:40-5:00 PM | | NVMF-302B-1: Are Ethernet Attached SSDs Happening? (NVMe-oF Track) | | Paper Presenters:
| NVMe-oF Ethernet SSDs Optimizing eBOF Solutions for Efficient Data Center
Ilker Cebeli, Sr Director Product Planning, Samsung SemiconductorBuilding a Better Solution with Native NVMe-oF SSDs
John Kloeppner, Principal Engineer / HW Architect, NetAppBenefits of Native NVMe-oF SSD
Balaji Venkateshwaran, Director Product Mgmt & Strategy, CiscoNative NVMe-oF SSD
Khurram Malik, Director of Product Marketing, MarvellStorage System Using NVMe over Fabric SSD-Based Ethernet JBOF
Woo Suk Chung, Storage Software Team Lead, SK hynix | Session Description:
| Ethernet-attached SSDs are one of the most interesting new developments in networked storage. Enabled by NVMe-oF, this new architecture removes the bandwidth limitation and lowers the cost of current JBOF designs. Today’s NVMe-oF Ethernet JBOF designs are typically architected with PCIe switches, Ethernet NICs and CPU complexes all there to convert NVMe-oF on Ethernet to NVMe on PCIe and back. Although widely successful to date this approach limits the NVMe SSD performance to the data rate capability of the CPU and NICs on its PCIe bus. It is also not cheap. This session will show how NVMe-oF Ethernet attached SSDs (eSSD) and Ethernet storage enclosures (eBOF) could address both these JBOF limitations.
| | About the Organizer/Moderator: | |
|