| Tuesday, December 4th |
Tuesday, December 4th
6:00-7:00 PM | | PCON-1: NVMe for the Rank Beginner (Pre-Conference Track) | | Speaker(s):
| Presenter: Mike Heumann, Managing Partner, G2M Communications | Session Description:
| Have you never worked with NVMe before and aren’t sure what all the terms mean? According to a recent ActualTech Media survey of IT professionals, you are in the majority – they found that, “Almost one-third (31%) of respondents have never even heard the term NVMe, and 54% say that they’ve heard of it but do not have experience with it.” This seminar is for you, and will give you the background needed to understand the presentations in the main conference. It will cover the reasons behind NVMe, its basic functionality, later additions to the standard, typical implementations, use cases and applications, resources, and future trends. Attendees will get a workbook they can use to create an NVMe strategy suited to their needs and application area.
| | About the Organizer/Moderator: | Speaker Bio: Mike Heumann is Managing Partner at G2M Communications, a market research firm focused on emerging technologies. He heads the development of research reports on NVMe, artificial intelligence, and machine learning. He was previously VP North American Sales for QLogic and VP Product Marketing and Alliances for Emulex. He has over 15 years experience in the storage industry with an emphasis on Fibre Channel and storage area networks. Mike earned an MS in Organizational and Industrial Psychology at Purdue University and a BSEE at the University of Wisconsin-Madison. He is the Conference Chair for the NVMe Developer Days event.
|
|
Tuesday, December 4th
7:00-8:00 PM | | PCON-2: 60 NVMe Tips and Tricks in 60 Minutes (Pre-Conference Track) | | Chairperson: Mike Heumann, Managing Partner, G2M Communications | Speaker(s):
| Presenter: Jonmichael Hands, Product Marketing Manager, Intel | Session Description:
| Users and developers alike have learned a lot about the practical aspects of NVMe in the past few years. Now they’ll share key tidbits with the rest of the world. We’ll have the best of tips, tricks, hints, warnings, gotchas, and plain old advice for everyone who is currently working or planning to work with NVMe. If you have practical wisdom to share, send it to us via email. If you just want to listen, come join us. If we don’t hit 60 in an hour, we will owe you something for next year! NVMe-oF and NVMe-MI also welcome. Remember – we want ideas that could send attendees away feeling that they already got enough from the conference to make the time, money, and general nuisance of traveling all worthwhile.
| | About the Organizer/Moderator: | Mike Heumann is Managing Partner at G2M Communications, a market research firm focused on emerging technologies. He heads the development of research reports on NVMe, artificial intelligence, and machine learning. He was previously VP North American Sales for QLogic and VP Product Marketing and Alliances for Emulex. He has over 15 years experience in the storage industry with an emphasis on Fibre Channel and storage area networks. Mike earned an MS in Organizational and Industrial Psychology at Purdue University and a BSEE at the University of Wisconsin-Madison. He is the Conference Chair for the NVMe Developer Days event.
Speaker Bio: Jonmichael Hands is a Senior Strategic Planner for Intel’s Non-Volatile Memory Solutions Group (NSG). He is currently the product line manager for data center SSDs, handling product marketing, marketing management, and strategic planning. He is also Co-chair of the Marketing Workgroup for NVM Express, Inc., the standards organization for NVMe. He has been very active in NVMe standards work, and has presented on NVMe at webinars and conferences, including Flash Memory Summit. He has 10 years experience in the technology industry, has written four articles, and has received a patent. He earned a BSEE from the Colorado School of Mines.
|
|
Tuesday, December 4th
8:00-9:00 PM | | PCON-3: SPDK Tutorial (Pre-Conference Track) | | Chairperson: Mike Heumann, Managing Partner, G2M Communications | Speaker(s):
| Presenter: Ben Walker, Storage Solutions Architect, Intel | Session Description:
| The Storage Performance Development Kit (SPDK) is an open-source set of tools and libraries for writing high performance, scalable, user-mode storage applications based on the NVMe interface. It achieves high performance by moving drivers into userspace, polling for completion instead of using interrupts, and avoiding locks in the I/O path. It includes a highly efficient NVMe driver for accessing an SSD from a user space application and a full block stack as a user space library. It also includes servers for common targets including NVMe-oF. Users will find it to be highly efficient, easy to integrate into their applications, and a good basis for production deployments. Releases of SPDK are available through GitHub, and the software is supported by an active Website (spdk.io) and an annual conference (see www.spdk.io/summit/us/2018/ for a description of the 2018 event).
| | About the Organizer/Moderator: | Mike Heumann is Managing Partner at G2M Communications, a market research firm focused on emerging technologies. He heads the development of research reports on NVMe, artificial intelligence, and machine learning. He was previously VP North American Sales for QLogic and VP Product Marketing and Alliances for Emulex. He has over 15 years experience in the storage industry with an emphasis on Fibre Channel and storage area networks. Mike earned an MS in Organizational and Industrial Psychology at Purdue University and a BSEE at the University of Wisconsin-Madison. He is the Conference Chair for the NVMe Developer Days event.
Speaker Bio: Ben Walker is a SPDK Core Maintainer at Intel, where he is the technical lead for the Storage Performance Development Kit (SPDK). As Technical Lead, he manages additions and revisions, reviews code, does performance analysis, and handles testing, code quality control, and documentation. He was previously the technical lead for Intel Rapid Storage Technology, where he implemented NVMe support. Before joining Intel, Ben worked on software projects ranging from particle physics simulation to high frequency securities trading. He earned a BS in physics from Arizona State University. He has presented on SPDK at many events, including SPDK Summit and Flash Memory Summit. He holds a patent in I/O queue management.
|
|
| Wednesday, December 5th |
Wednesday, December 5th
10:00-12:00 PM | | Plenary 1: NVMe Today - Where It Is and Where It’s Heading (Plenaries Track) | | Organizer + Speaker: Peter Onufryk, Fellow Data Center Solutions BU, MicrochipChairperson: Cameron Brett, Director SSD Product Marketing, TOSHIBA AMERICA ELECTRONIC COMPONENTSCo-Organizer: Praveen Midha, Director Product Management/Strategic Marketing, Marvell | Paper Presenters:
| NVMe - Delivering Continued Innovation for Storage
Jonmichael Hands, Product Marketing Manager, IntelNVMe Ecosystem Market Report
Mike Heumann, Managing Partner, G2M CommunicationsNVMe over Fabrics (NVMe-oF): High-Speed, Low-Latency Shared Storage
Rob Davis, VP Storage Technology, MellanoxThe NVMe Management Interface (NVMe-MI): Overview and New Developments
Peter Onufryk, Fellow Data Center Solutions BU, Microchip | Session Description:
| NVM Express (NVMe) has quickly become the most popular interface for storage devices. Its advantages over disk interfaces (SAS and SATA) are obvious: higher speed, lower power, greater ruggedness, lower latency, and more flexibility through multiple command queues. Recent versions of NVMe have added more features needed in large storage systems based on solid-state devices. Also supplemental standards have been added, including the management interface NVMe-MI and the networking (fabric) interface NVMe-oF. Market research shows rapid acceptance of NVMe in enterprise storage, computers, high-performance computing, financial systems, embedded systems, mil/aero, automotive, and even mobile devices.
| | About the Organizer/Moderator: | Peter Onufryk is Fellow in the Data Center Solutions business unit at Microchip, where he is responsible for storage product architecture. He has been very active in NVMe standardization as an NVMe Board Member and NVMe Management Workgroup Chair. He has been a featured speaker at many events on NVMe and NVMe-MI, including Flash Memory Summit. He holds over 40 patents in interfaces and communications and has written several published articles. He was previously Director of Engineering at Integrated Device Technology (IDT) and a research staff member at AT&T Bell Labs. He earned a PhD in electrical and computer engineering from Rutgers University and an MSEE from Purdue University.
Cameron Brett is the Director of Enterprise SSD Marketing at Toshiba America Electronic Components, Inc. where he manages a team of product line managers to drive product strategy and revenue growth. Cameron have over 18 years of product marketing and management experience in storage technology and has previously held managerial positions at QLogic, PMC-Sierra, Broadcom and Adaptec. Throughout his career in high-tech product marketing, he has focused on storage for enterprise and small/medium business servers and worked to bring new generations of storage technology to market. His area of expertise includes Flash/SSD storage, virtualization, convergence and cloud technologies.
Praveen Midha is Director Product Management and Strategic Marketing at Marvell Semiconductor, where he focuses on storage solutions including FC-NVMe. He has successfully brought several new technologies (including 32G Fibre Channel adapters) in enterprise storage, networking, and virtualization to market and directed full lifecycle for products with multi-hundred million dollar revenues. He has over 25 years experience in the technology industry. He earned an MBA from Santa Clara University and a Bachelor in Technology in electronics and communications engineering from the Indian Institute of Technology Varanasi. He has been active in INCITS T11 (Fibre Channel) and has presented on NVMe at Flash Memory Summit.
|
|
Wednesday, December 5th
2:00-3:15 PM | | APPL-101: Moving Existing Applications to NVMe/NVMe-oF (Applications Track) | | Organizer: Howard Marks, Chief Scientist, DeepStorageChairperson: Frederic Van Haren, Sr Analyst/Practice Lead, Evaluator GroupOrganizer: Radjendirane Codandaramane, Sr. Manager, Applications Engineering, Microsemi | Paper Presenters:
| Selecting the Right Architecture for an All-Flash NVMe Solution
Mike Scriber, Sr Director Server Solutions Management, SuperMicroPCIe Networked Flash Storage
Peter Onufryk, Fellow Data Center Solutions BU, MicrochipIs NVMe Right for Military/Industrial Applications?
Chris Budd, Director Engineering, SMART High Reliability Solutions | Session Description:
| Moving existing applications from disk interfaces to NVMe is not as simple as it may sound. Many applications use disk interface features that do not exist in NVMe, and obviously none takes advantage of NVMe-only features. Designers must also be careful about using system software that may not recognize NVMe at all or where the support may be only rudimentary. The advantages of NVMe in some applications, such as embedded, may be offset by its high power consumption and limited support.
| | About the Organizer/Moderator: | Howard Marks has been writing, speaking, and consulting about enterprise technology for over thirty years. As a consultant, he has designed storage, server, and network infrastructures for organizations such as The State University of New York (Purchase) , BBDO Worldwide, and the Foxwoods Resort Casino. He also operates an independent laboratory (DeepStorage) which tests storage products for both vendors and magazines. He started testing and reviewing products at PC Magazine in the late 1980s and has written hundreds of articles and product reviews for such media as Network World, Network Computing, and InformationWeek. A top rated speaker at industry events, he has spoken at Storage Decisions, Interop, Microsoft’s TechEd, and Flash Memory Summit. He has also developed training programs for organizations such as JP Morgan and American Express.
Frederic Van Haren joined Evaluator Group in 2017 to lead the HPC/AI practice. With over 20 years of experience in high technology, Frederic is known for his insight into the HPC and AI markets from his hands-on experience leading research and development teams. He specializes in working with engineering teams from technology vendors, providing feedback on projects. He is particularly interested in the telecom and speech markets, having spent 10 years at Nuance Communications building large HPC environments from the ground up. He is frequently invited to speak at events to provide his vision of the HPC, AI/ML, IoT, and storage markets. Frederic has also served as the president of a variety of user groups promoting innovative technology. He has published several technical articles and speaks four languages.
Radjendirane (Radj) Codandaramane is Sr Manager Applications Engineering in the Data Center Solutions Business Unit of Microsemi. He manages the customer support team for NVMe controllers and NVRAM drives, including related hardware, firmware, and software. He has previous experience at PMC-Sierra and Agilent Technologies. He has strong technical skills in SoCs, embedded firmware, drivers, and major storage protocols and protocol controllers. Radj has contributed to the development of enterprise storage products for the last 17 years. He earned a BS in electronics and communications engineering at Pondicherry University (India). He has been a frequent speaker at past Flash Memory Summits.
|
|
Wednesday, December 5th
2:00-3:15 PM | | HRDW-101: Designing NVMe/NVMe-oF Controllers (Hardware Track) | | Chairperson: Deepankar Das, CTO, Sureline SystemsOrganizer: Rakesh Cheerla, Solution Planner, IntelCo-Organizer: Sean Gibb, VP Software, Eideticom | Paper Presenters:
| Design Considerations for M.2 NVMe SSD Adapters
Mohammad Darwish, CEO, Aplicata TechnologiesUltra Low Latency NVMe-oF Controller Design
Swati Chawdhary, Senior Manager, SamsungIncreasing System Throughput by Adding Compute Power to NVMe Controllers
Scott Shadley, VP Marketing, NGD Systems | Session Description:
| NVMe controller design is changing rapidly as vendors try to gain market share in a rapidly expanding area. New features include direct control of the flash in system software (open-channel SSD), improved support for streams and virtualization, programmability, addition of local memory to avoid contention for system resources, support for NVMe-oF and fabrics, higher speeds, data integrity and security features, support for PCIe Gen 4, and the use of machine learning and artificial intelligence techniques. There are obvious tradeoffs here among such factors as cost, lifetime, flexibility, throughput, latency, and ease of use.
| | About the Organizer/Moderator: | Deepankar Das is CTO of Sureline Systems, driving the leading edge in application mobility to allow machines, VMs, and applications to move seamlessly between physical, virtual, and cloud infrastructure. Before joining Sureline, he was Head of Engineering for the EMC Data Domain file system where he delivered the next generation Data Domain Data Protection in the Cloud products. He was previously Head of Software Engineering at MRAM startup Avalanche Technology, where he was in charge of creating software for a super-high-performance all-flash storage array, including Block/File Storage, Kernel/ Platform, HA/Clustering, Flash Management, SSD Firmware, and GUI. He has also been Head of Software Engineering at Violin Memory, where he was engineering leader for the overall Violin Software, including high performance vMOS stack, Violin-Symantec Data Management stack, OEM/Platform software, Target Device Drivers, Violin Memory Array Device Drivers, Virtualization, User Interface, and Release Engineering. He has also worked for EMC, Panasas, and Sun Microsystems. He earned a Master’s degree in computer science from Andhra University (India).
Rakesh Cheerla is currently Product Manager for Storage Solutions at Xilinx, where he focuses on data center design, hardware management, and open storage standards. He was previously Sr Director Products at CNEX Laboratories, a startup developing flash controller chips, where he was in charge of determining customer requirements and defining products to meet them. Before joining CNEX, he worked at SMART Modular Systems, LSI, and Extreme Networks. He has focused on developing product requirements, defining features, and managing engineering teams. He holds an MBA from Columbia University and an MSEE from Arizona State University. He is a member of the Conference Advisory Board for both Flash Memory Summit and NVMe Developer Days.
Sean Gibb is a software professional with over 20 years of experience initiating and leading complex software development projects. Before founding Eideticom, he was Lead Software Engineer at PMC-Sierra where he performed software system and algorithm development of LDPC FEC/ECC for enterprise-grade SSD controllers. Before joining PMC, he was a founder and Director of Software at Rad3 Communications, a Senior Staff Engineer at Nextwave Wireless, and a digital IC designer at SiWorks. He is an expert on algorithm acceleration and implementation of algorithms on multi-core processors and other programmable platforms. Sean is an author of several novel patents on FFT architectures and holds a PhD in electrical engineering from the University of Calgary where he specialized in the automatic generation of application specific processors.
|
|
Wednesday, December 5th
2:00-3:15 PM | | NETW-101: Discussion Group on Networking (Networking Track) | | Chairperson: J Metz, Office of the CTO/Board Member, Cisco SystemsOrganizer: Bryan Cowger, VP Sales/Marketing, Kazan NetworksCo-Organizer: Vishal Shukla, Director, WW Ethernet Switch Technology, Mellanox | Panel Members:
| Panelist: Sagi Grimberg, Principal Architect, Lightbits LabsPanelist: Jeff Shao, Director Ethernet Alliances, MellanoxPanelist: Bob Dugan, Director Engineering, Chelsio Communications | Session Description:
| Networking flash storage is a complex business. NVMe-oF provides a basic framework, but there is still much work for designers to do. They must decide on a fabric mechanism, and must buy or develop software that will manage it. They also need an addressing mechanism, switches, and ways to avoid long delays that could produce extremely high tail latency. Networks always introduce delay, but congestion or noisy neighbors can make the amount intolerable. Faster networks can help increase throughput, but may require expensive components and test equipment. Designers may also need to add local resources to reduce delays and avoid competition. Obviously, a wide range of solutions exist. Choosing among them depends on basic tradeoffs such as cost vs. performance, as well as more complex ones such as system complexity vs. maintainability.
| | About the Organizer/Moderator: | J Metz is currently an R&D Engineer for the Office of the CTO in Cisco’s Compute and Server Group, where he focuses on directions for storage strategy. He is an award-winning public speaker, author, and contributor to industry trade publications, blogs, webinars, and conferences. He has been a leader in developing industry standards, with membership on the Board of Directors for the Fibre Channel Industry Association (FCIA), Storage Networking Industry Association (SNIA), and the Non-Volatile Memory Express (NVMe) Promoter’s Board. J has previous experience with QLogic and Apple. He earned his PhD from the University of Georgia. J has been a speaker, panelist, and chairperson in well-received sessions at several past Flash Memory Summits.
Bryan Cowger is VP Sales/Marketing at Kazan Networks, a startup focused on developing ASICs for NVMe-oF and NVMe TCP solutions. He has almost 30 years experience in the technology industry with a heavy focus on storage networking. Before joining Kazan Networks, he was VP Sales and VP Marketing at Emulex, a leader in Fibre Channel networking. He also worked at Sierra Logic and Agilent Technologies. He earned a BSEE from UC San Diego. He has presented at Flash Memory Summit and is a member of the Conference Advisory Board for NVMe Developer Days. He holds 4 patents in storage controller architectures.
Vishal Shukla is Sr Director Systems Engineering at NVXL Technology, a maker of compute acceleration products based on machine learning. A thought leader in networking, cloud, and deep learning solutions, he has written multiple papers, authored books, and obtained over 15 patents. He has over 10 years proven experience in developing and leading global teams in functions such as technical sales, solutions architecture, technical alliances, and customer support. He has worked for Mellanox, IBM, Blade Network Technologies, Cisco, and Nortel. He earned an MBA from Duke University and a BS in Information Technology from Uttar Pradesh Technical University (India).
|
|
Wednesday, December 5th
3:30-4:45 PM | | APPL-102: Developing New Applications (Applications Track) | | Chairperson: David McIntyre, Principal Consultant, DS McIntyre ConsultingCo-Organizer: Deepankar Das, CTO, Sureline SystemsOrganizer: Scott Shadley, VP Marketing, NGD Systems | Paper Presenters:
| Computational Storage in NVMe-Based System
Sean Gibb, VP Software, EideticomCreating Mobile NVMe Applications with SDExpress
Parag Beeraka, Sr Manager Product Management, Western DigitalUnderstand the Use Cases Driving NVMe-over-Fabrics Adoption
Julie Herd, Director Technical Marketing, E8 Storage | Session Description:
| New applications require careful analysis. NVMe is still in its early stages, so designers must be aware of what services the specification provides and what they must still add on top. Designers need to be fully aware of open-source software such as SPDK which can provide important assistance in creating efficient systems. Testing remains an issue, since support is still in its infancy and equipment may be difficult to obtain.
| | About the Organizer/Moderator: | David McIntyre is Principal Consultant at DS McIntyre Consulting, where he focuses on machine learning solutions, as well as programmable logic and GPU applications in the data center. He previously led the compute and storage business for Altera (now part of Intel). He has also held leadership positions at IBM, Fairchild, and startup Transmeta where he marketed and led the development of comprehensive product portfolios for high performance computing, enterprise networking, and enterprise storage. He earned MSEE and BSEE degrees at Ohio University and an MBA at San Jose State University. David is a regular speaker at networking and data center infrastructure conferences, including Flash Memory Summit.
Deepankar Das is CTO of Sureline Systems, driving the leading edge in application mobility to allow machines, VMs, and applications to move seamlessly between physical, virtual, and cloud infrastructure. Before joining Sureline, he was Head of Engineering for the EMC Data Domain file system where he delivered the next generation Data Domain Data Protection in the Cloud products. He was previously Head of Software Engineering at MRAM startup Avalanche Technology, where he was in charge of creating software for a super-high-performance all-flash storage array, including Block/File Storage, Kernel/ Platform, HA/Clustering, Flash Management, SSD Firmware, and GUI. He has also been Head of Software Engineering at Violin Memory, where he was engineering leader for the overall Violin Software, including high performance vMOS stack, Violin-Symantec Data Management stack, OEM/Platform software, Target Device Drivers, Violin Memory Array Device Drivers, Virtualization, User Interface, and Release Engineering. He has also worked for EMC, Panasas, and Sun Microsystems. He earned a Master’s degree in computer science from Andhra University (India).
Scott Shadley is VP Marketing at NGD Systems, a developer of computational storage. He focuses on brand, go-to-market, and product management and development, as well as customer adoption, acquisition, and support. He was previously Principal Technologist at Micron, where he focused on future technologies such as form factors, interfaces, and connectivity options with a particular emphasis on giving customers more freedom with their uses. He has spent over 20 years in the semiconductor and storage industries, and his efforts have been important in developing and marketing products with over $300M revenue. He has also been active as a conference presenter and organizer, particularly at Flash Memory Summit. He earned an MBA from the University of Phoenix and a BS in Semiconductor Device Physics from Boise State University (ID).
|
|
Wednesday, December 5th
3:30-4:45 PM | | HRDW-102: Designing NVMe/NVMe-oF Storage Systems (Hardware Track) | | Chairperson: Murali Iyer, Chief Engineer for NVMe Storage Solutions, IBM Cognitive Systems GroupOrganizer: David Black, Distinguished Engineer, Dell EMC | Paper Presenters:
| Improving Cassandra's Performance with Docker and NVMe
Kais Belgaied, Storage & Servers Division CTO, Sanmina CorpDavid Paulsen,
,
Sanmina Corp
The Design of Key-Value Store for Data Acquisition Systems
Maciej Maciejewski, Software Architect, IntelJakub Radtke,
,
Intel
Creating Efficient Storage Networks with Ethernet and NVMe-oF
J Metz, Office of the CTO/Board Member, Cisco Systems | Session Description:
| NVMe storage systems require careful planning. NVMe drives require a steady stream of high-frequency data to operate smoothly, so links must be fast and overhead must often be offloaded. Systems can easily run out of processing power as the number of drives increases. System utilities are still in their infancy, so designers need to look for packages that can handle the full capabilities of NVMe. That includes programs for system-level tasks such as compression, deduplication, and mirroring. The easiest improvements and the fastest to achieve involve applications that have not undergone many years of optimization, such as NoSQL databases, AI and ML, VR and AR, and real-time analytics.
| | About the Organizer/Moderator: | Murali Iyer is Chief Engineer for NVMe Storage Solutions at IBM Cognitive Systems Group, where he focuses on developing new NVMe products including devices and direct (PCIe) and fabrics-attached enclosures.. He works on defining NVMe product features, device form factors, roadmaps, and feature sets suitable for AIX, Linux, and IBMi operating systems for POWER and OpenPowerŽ systems. He provides technical updates to engineering teams across different groups within IBM as well as to customers. He is the primary contact from IBM for NVMExpress.org. He has over 30 years of industry experience including at 20 years at IBM.. For the past nine years, he has focused on storage RAID adapter and NVMe-based design and development. He holds 12 patents in the storage area.
David L. Black is a senior distinguished engineer in Dell EMC's Office of the CTO, where he works on technology and standards with a long-term (3-5 year) focus and shorter term applications to products. He has worked on many strategic technology projects and standards efforts, including the iSCSI protocol, Fibre Channel security, and NVMe over Fabrics. He has been active in many standards organizations, including IETF (e.g., iSCSI), SNIA, T10 (SCSI), and T11 (Fibre Channel). He holds 33 patents and has presented at many conferences, including Flash Memory Summit and Storage Networking World. He earned a PhD and MS in computer science from Carnegie-Mellon University and a BSE in computer science and engineering (summa cum laude) from the University of Pennsylvania.
|
|
Wednesday, December 5th
3:30-4:45 PM | | SOFT-102: Development Software (Software Track Track) | | Chairperson: Uma Parepalli, Sr Manager, MarvellCo-Organizer: Allen Samuels, Engineering Fellow, Western Digital | Paper Presenters:
| Distributed Endpoint Management: An NVMe-oF Scale-Out Management Solution
Phil Cayton, Sr Staff Engineer, IntelImplementing End-to-End NVMe in a High-Performance Flash Memory System
Chris Dennett, Sr Technical Staff Member, IBMSPDK's NVMe over Fabrics Software Target
Ben Walker, Storage Solutions Architect, Intel | Session Description:
| Software is obviously the biggest issue in developing NVMe applications. NVMe itself involves only the well-established PCIe bus as its hardware base. But the software must all be built from the ground up. That includes drivers, utilities, management packages, RAID and high-availability software, data migration software, software-defined storage, and other programs at all levels. Some is available open-source, but far more must be developed in-house or acquired from a limited number of sources. The open-source SPDK is an example of the types of tools developers need, particularly if they want to match the mature tools available for disk interfaces.
| | About the Organizer/Moderator: | Uma Parepalli currently heads firmware engineering in Marvell's ARM Processor Business Unit. Uma has 27 years experience in the server and storage industries with additional experience in embedded systems, consumer electronics. and the aerospace industry. Before joining Marvell, Uma headed engineering and product development teams, and worked in senior level architecture and management roles from director to VP/head of engineering. He has worked for Western Digital, SK hynix, Avago, Dell EMC, Intel, Wipro, and other industry leaders. He has been active in many important standards groups, including NVM Express, UEFI, and ACPI, and has been a presenter, chairperson, and Conference Advisory Board member for the Flash Memory Summit. Uma earned a computer engineering degree from Mysore University (India).
Allen Samuels is an Engineering Fellow at Western Digital, where he helps direct the CTO office’s research into software and systems. He is a long-time leader in the open-source software community and has spoken at many open-source related events. He is a member of the Ceph Advisory Board and is heavily involved in the evolution of that critical open source technology in software-defined storage. Before joining Western Digital, he was a Chief Architect at Weitek and Citrix, and founded several companies including AMKAR Consulting, Orbital Data, and Cirtas Systems. He is a frequent speaker at conferences and other events, including Flash Memory Summit. He earned a BSEE from Rice University.
|
|
| Thursday, December 6th |
Thursday, December 6th
9:00-10:00 AM | | Plenary 2: Optimization of NVMe Systems: Performance, Power, Cost, and Latency (Plenaries Track) | | Chairperson: Sean Gibb, VP Software, EideticomOrganizer: Allan Cantle, CEO/Founder, Nallatech | Panel Members:
| Panelist: Fazil Osman, Distinguished Engineer, BroadcomPanelist: Benoit Ganne, Chief Architect, KalrayPanelist: Zhiping Yang, Hardware Engineer, GooglePanelist: Mike Heumann, Managing Partner, G2M Communications | Session Description:
| As with most systems, NVM-based designs generally need an optimization effort to produce high performance levels. Also systems, when tested, often do not meet specifications or requirements, provide the performance users expect, or perform as well as competitive products. Basic approaches to optimization include developing faster hardware or adding hardware acceleration, optimizing the software, and making minor hardware additions that can solve short-term problems. The trade-offs are generally obvious. They include design time and time available to make the revision, cost, maintainability, long-term compatibility with standards, and the amount of improvement needed. Storage systems are often highly complex, and improvements in one area may not have much overall effect because of side effects and the exposure of other performance-limiting conditions. On the hardware side, adding everything from mezzanine cards through FPGAs and co-processors to upgraded networks and computational storage can all make a difference. On the software side, optimized stacks may make a significant difference as can reducing system calls, using polling instead of interrupts, and changing flash assignment methods to increase parallelism and work better with specific applications (so-called I/O determinism).
| | About the Organizer/Moderator: | Sean Gibb is a software professional with over 20 years of experience initiating and leading complex software development projects. Before founding Eideticom, he was Lead Software Engineer at PMC-Sierra where he performed software system and algorithm development of LDPC FEC/ECC for enterprise-grade SSD controllers. Before joining PMC, he was a founder and Director of Software at Rad3 Communications, a Senior Staff Engineer at Nextwave Wireless, and a digital IC designer at SiWorks. He is an expert on algorithm acceleration and implementation of algorithms on multi-core processors and other programmable platforms. Sean is an author of several novel patents on FFT architectures and holds a PhD in electrical engineering from the University of Calgary where he specialized in the automatic generation of application specific processors.
Allan Cantle is CTO/Founder at Nallatech, a Molex company. He focuses on developing FPGA solutions in defense, high-performance computing, and server applications. He also presents on FPGA computing at conferences around the world and extends Nallatech’s solutions to new applications. He has over 35 years experience in the technology industry, including 25 years at Nallatech. He earned a degree in electrical and electronics engineering at the University of Plymouth (UK) and a master’s in corporate leadership at Napier University. Cantle holds 3 patents and has published several technical articles.
|
|
Thursday, December 6th
2:00-3:15 PM | | NETW-201: NVMe-oF Reference Designs (Networking Track) | | Organizer + Speaker: John Kim, Director Storage Marketing, MellanoxOrganizer: Weafon Tsao, VP, AccelStorOrganizer: Ziv Serlin, VP System Architecture, E8 StorageChairperson: Vishal Shukla, Director, WW Ethernet Switch Technology, Mellanox | Paper Presenters:
| NVMe-oF JBOF with Acceleration Capabilities
Rakesh Cheerla, Solution Planner, IntelGildas Genest,
,
BittWare
End-to-End NVMe-oF Design
John Kim, Director Storage Marketing, Mellanox | Session Description:
| There are many choices of transports to use in an NVMe-oF network. The most obvious one is high-speed Ethernet due to its ubiquity, low-cost components, long experience, and familiarity in all installations. However, Ethernet was not intended for such purposes and its throughput is generally quite low, requiring either extremely high data rates or external transport offloading to compensate. Other choices include InfiniBand, Fibre Channel, PCIe, and new alternatives such as Gen-V. Typical selection issues are cost, bandwidth, ecosystem, familiarity, security, power consumption, and support.
| | About the Organizer/Moderator: | John F. Kim is Director of Storage Marketing at Mellanox Technologies, where he helps storage customers and vendors benefit from high performance interconnects and smart offloads, including RDMA (Remote Direct Memory Access). He is a frequent conference participant, including several past Flash Memory Summits, and a frequent blogger on storage and networking topics. He is also chair of the Ethernet Storage Forum. Before joining Mellanox, he created storage solutions and alliances at NetApp and EMC. He has a BA from Harvard University. Follow him on Twitter: @Tier1Storage
Weafon Tsao is VP R&DS AccelStor, a startup focused on a software-defined approach to flash-based storage solutions. He is the lead inventor of AccelStor’s FlexiRemap software technology, which won the award for Most Innovative Flash Memory Technology at Flash Memory Summit 2016. Dr. Tsao was previously Deputy Division Director of the Cloud Computing Center for Mobile Applications at ITRI (Industrial Technology Research Institute), tasked with building a flash memory storage system for cloud computing and power savings. He has over 17 years experience in telecommunications, networking, and storage. Dr. Tsao earned his MS and PhD in Computer Science from National Chiao Tung University and did postdoctoral research at Lawrence Berkeley Laboratory.
Ziv Serlin is VP Systems Architecture and Co-Founder at E8 Storage, a maker of high-performance storage appliances. He specializes in the architecture, design, prototyping, and integration of complex storage and networking systems. Before co-founding E8 Storage, he was a system architect at Primary Data and a System Architect/Hardware Manager at IBM (where he helped develop XIV, a high-end grid-scale storage system). He earned an MSEE and MBA at Tel Aviv University and a BSc in computer engineering cum laude from the Technion (Israel Institute of Technology). He holds a key storage patent on reducing data loss from his work at IBM and has presented at several conferences, including Flash Memory Summit.
Vishal Shukla is Sr Director Systems Engineering at NVXL Technology, a maker of compute acceleration products based on machine learning. A thought leader in networking, cloud, and deep learning solutions, he has written multiple papers, authored books, and obtained over 15 granted patents. He has over 10 years proven experience in developing and leading global teams in functions such as technical sales, solutions architecture, technical alliances, and customer support. He has worked for IBM, Blade Network Technologies, Cisco, and Nortel. He earned an MBA from Duke University and a BS in Information Technology from Uttar Pradesh Technical University (India).
|
|
Thursday, December 6th
2:00-3:15 PM | | SOFT-201: Discussion Group on Software (Software Track Track) | | Chairperson: Alan Weckel, CEO/Founder, 650 GroupOrganizer: Dave Minturn, Principal Engineer, IntelOrganizer + Speaker: Max Kolomyeytsev, Director of Product Management, Starwind SoftwareOrganizer: Uma Parepelli, Sr Manager, MarvellOrganizer: Kais Belgaied, Storage & Servers Division CTO, Sanmina Corp | Paper Presenters:
| Bare Metal Performance for Software Defined Storage with NVMe
Phil White, CTO, Scale ComputingDeveloping High-Performance, Low-Latency Applications with NVMe-oF
Anton Kolomyeytsev, Co-Founder/CEO/ Chief Architect, Starwind SoftwareMax Kolomyeytsev,
,
Starwind Software
Yuriy Khokhlov,
,
StarWind Software
| Session Description:
| NVMe application developers will need a lot of software support to work efficiently Packages like the open-source SPDK are an important part of the answer because they are readily available, well-supported, and easy to use. Other areas will include high-level packages that allow developers to specify their needs in an easily understood manner, utilities packages to cover specific applications and other needs, and packages that will allow developers to use the latest constructs such as containers and security suites. Extensive software support is a key aspect to creating a mature development environment and making standards like NVMe and NVMe-oF achieve wide usage. Clearly, open-source development efforts must play a major role in making this feasible.
| | About the Organizer/Moderator: | Alan Weckel is Technology Analyst/Co-Founder at 650 Group, where he is in charge of Ethernet switch research and new areas such as SDN forecasting and WAN optimization. He has written many articles for the trade and technical press, and is frequently quoted in such leading publications as Bloomberg, Businessweek, Forbes, Network World, and the Wall Street Journal. Before co-founding 650 Group, he was VP/analyst at Dell’Oro Group and had engineering and software development experience at Raytheon, General Electric Power Systems, and Cisco. He holds a BSEE and an MS in Management from Rensselaer Polytechnic Institute.
Dave Minturn is a Principal Engineer and Network SSD Storage Architect at Intel. He is one of the principal architects of the NVMe over Fabrics (NVMe-oF) specification and has developed architectures and optimizations for NVMe PCIe SSDs in networked storage systems. He has also worked on RDMA storage protocols for NVM and software stack optimizations for SSDs. He holds patents in networking and has presented at many conferences, including Intel Developer Forum and Flash Memory Summit as well as at events sponsored by SNIA, USENIX, and Open Fabrics Alliance. He was previously a Sr Network Architect at Intel.
Max Kolomyeytsev is a Director of Product Management at StarWind Software, a developer of Software-Defined Storage for hyperconverged infrastructures. He focuses on hyperconverged and VDI reference architectures used to improve infrastructure resiliency and performance. An IT professional for over 12 years, he has experience in performance tuning, QA, solutions architecting, and engineering. He is an active commentator on Spiceworks as Max (StarWind), and @Max_Schmeisser on Twitter. He has participated in many webinars as well as at events such as MangoCon and TechField Day.
Uma Parepalli currently heads firmware engineering in Marvell's ARM Processor Business Unit. Uma has 27 years experience in the server and storage industries with additional experience in embedded systems, consumer electronics. and the aerospace industry. Before joining Marvell, Uma headed engineering and product development teams, and worked in senior level architecture and management roles from director to VP/head of engineering. He has worked for Western Digital, SK hynix, Avago, Dell EMC, Intel, Wipro, and other industry leaders. He has been active in many important standards groups, including NVM Express, UEFI, and ACPI, and has been a presenter, chairperson, and Conference Advisory Board member for the Flash Memory Summit. Uma earned a computer engineering degree from Mysore University (India).
Kais Belgaied is currently CTO for Sanmina's Storage and Servers division, where he is driving the company's growth strategy around NVMe. A prolific innovator with over 50 patents in such areas as cache management, scalability, and hardware offload, he was previously an architect and team leader at Nimble Storage, where he defined the control plane scale-out and availability features for hybrid and all-flash array clusters. He has also worked for Coraid, VMware, and Sun Microsystems. He has published articles and given or contributed to presentations at events including Flash Memory Summit. He earned an MS and a postgraduate degree in computer science and applied mathematics from the Polytechnique Institute of Grenoble (France), and an MBA from Columbia.
|
|
Thursday, December 6th
2:00-3:15 PM | | STDS-201: Discussion Group on New Specification Features (Standards Track) | | Chairperson: Praveen Midha, Director Product Management/Strategic Marketing, MarvellOrganizer: David Woolf, Sr Engineer Datacenter Technology, UNH-IOL | Panel Members:
| Panelist: Ross Stenfort, Hardware Systems Engineer, FacebookPanelist: Mark Carlson, Principal Engineer, Industry Standards, Toshiba MemoryPanelist: Rob Davis, VP Storage Technology, MellanoxPanelist: Curtis Ballard, Storage Technologist, HPEPanelist: Sagi Grimberg, Principal Architect, Lightbits Labs | Session Description:
| Many new features have been added to NVMe with recent versions. Many others are waiting in the queue for possible inclusion later. Some are intended for large-scale operations, others are meant for specific application areas, and still other represent requirements discovered during systems implementations. Only through experience and user comments can the NVMe specification reach the maturity required for data centers, clouds, and other applications. New features for Version 1.3 include directives, SR-IOV support, sanitize, boot partitions, and self-test.
| | About the Organizer/Moderator: | Praveen Midha is Director Product Management and Strategic Marketing at Marvell Semiconductor, where he focuses on storage solutions including FC-NVMe. He has successfully brought several new technologies (including 32G Fibre Channel adapters) in enterprise storage, networking, and virtualization to market and directed full lifecycle for products with multi-hundred million dollar revenues. He has over 25 years experience in the technology industry. He earned an MBA from Santa Clara University and a Bachelor in Technology in electronics and communications engineering from the Indian Institute of Technology Varanasi. He has been active in INCITS T11 (Fibre Channel) and has presented on NVMe at Flash Memory Summit.
David Woolf is Senior Engineer, Datacenter Technologies at the University of New Hampshire InterOperability Laboratory (UNH-IOL), where he coordinates the NVMe integrators list and plugfests. He has developed dozens of industry-reviewed test procedures and implementations as part of the team that has grown the UNH-IOL into a world-class center for interoperability and conformance testing. David has also helped organize many industry interoperability test events at both at the UNH-IOL facility and off-site locations. He has also presented webinars on NVMe and presented on NVMe testing and interoperability at many conferences including Flash Memory Summit and Open Compute Summit. He earned a BSEE at the University of New Hampshire.
|
|
Thursday, December 6th
3:30-4:45 PM | | INDT-202. Where Will NVMe Be 10 Years from Now and How Did We Get There? (Industry Trends Track) | | Organizer + Chairperson: Jean S. Bozman, VP/Principal Analyst, Hurwitz & AssociatesCo-Organizer: Sean Gibb, VP Software, Eideticom | Panel Members:
| Panelist: Howard Marks, Chief Scientist, DeepStoragePanelist: Allen Samuels, Engineering Fellow, Western DigitalPanelist: David Black, Distinguished Engineer, Dell EMCPanelist: Mark Webb, President, MKW Ventures | Session Description:
| What can we expect by 2025? How large and how mature will the market be? Which applications and which customers will be most prominent? Which areas will get the most attention? What features will be widely used and which will be largely forgotten? What will be the major milestones along the way? What will the competitive landscape be? Panelists will take out their crystal balls and try to look backward in time.
| | About the Organizer/Moderator: | Jean S. Bozman is Vice President and Principal Analyst at Hurwitz and Associates, where she covers data center infrastructure, cloud infrastructure, server and storage technology, and software-defined infrastructure (SDI). Before joining Hurwitz & Associates in 2016, she was Senior Product Marketing Manager at SanDisk, where she drove the discussion of enterprise workloads. Bozman has more than 20 years of experience covering the worldwide markets for operating environments, servers, and the workloads that run on servers. She was Research Vice President of IDC’s Worldwide Server Group from 2002-2013. She has been widely quoted in the press and in online publications, such as Bloomberg, CNET, eWeek, Reuters and TechTarget. Ms. Bozman holds a B.S. degree from the State University of New York (SUNY) at Stony Brook, and a master's degree from Stanford University
Sean Gibb is a software professional with over 20 years of experience initiating and leading complex software development projects. Before founding Eideticom, he was Lead Software Engineer at PMC-Sierra where he performed software system and algorithm development of LDPC FEC/ECC for enterprise-grade SSD controllers. Before joining PMC, he was a founder and Director of Software at Rad3 Communications, a Senior Staff Engineer at Nextwave Wireless, and a digital IC designer at SiWorks. He is an expert on algorithm acceleration and implementation of algorithms on multi-core processors and other programmable platforms. Sean is an author of several novel patents on FFT architectures and holds a PhD in electrical engineering from the University of Calgary where he specialized in the automatic generation of application specific processors.
|
|
Thursday, December 6th
3:30-4:45 PM | | NETW-202: NVMe-oF Transports (Networking Track) | | Chairperson: Chris DePuy, Research Analyst/Co-Founder, 650 GroupOrganizer: Vishal Shukla, Director, WW Ethernet Switch Technology, Mellanox | Panel Members:
| Panelist: J Metz, Office of the CTO/Board Member, Cisco SystemsPanelist: Scott Taylor, Director Software Engineering, GigaIOPanelist: Rob Davis, VP Storage Technology, MellanoxPanelist: Praveen Midha, Director Product Management/Strategic Marketing, MarvellPanelist: Michael Krause, Lead Architect, Gen Z Consortium | Session Description:
| NVMe-oF is a very complex specification and is quite difficult to use in practice. Several reference designs are now available to guide implementers on how to make everything work, including hardware, software, and networking. A detailed perusal of typical examples provides a basic grounding in system design as well as showing where potential problems lie and choices have to be made. The emphasis will be on problem definition, hardware design, software design, network design, system integration, testing, and performance analysis.
| | About the Organizer/Moderator: | Chris DePuy is Research Analyst/Co-Founder at 650 Group, a leading market intelligence research firm for communications, data center, and cloud markets. He was previously a Vice-President at Dell’Oro Group, responsible for the Carrier IP Telephony, Enterprise Edge, Wireless LAN, Wireless Packet Core, and Storage market research programs. He has over 20 years of financial analysis, business analysis, and engineering experience. He has been a consultant with TS Cap and a research analyst covering software, communications, and Internet with Bowman Capital and Morgan Stanley. He was named the top ranking Equity Research Analyst in the data networking sector by Institutional Investor and Greenwich Research Survey. The co-author of a book, he holds a Masters in engineering from Cornell and a BS in engineering from Union College.
Vishal Shukla is Sr Director Systems Engineering at NVXL Technology, a maker of compute acceleration products based on machine learning. A thought leader in networking, cloud, and deep learning solutions, he has written multiple papers, authored books, and obtained over 15 granted patents. He has over 10 years proven experience in developing and leading global teams in functions such as technical sales, solutions architecture, technical alliances, and customer support. He has worked for IBM, Blade Network Technologies, Cisco, and Nortel. He earned an MBA from Duke University and a BS in Information Technology from Uttar Pradesh Technical University (India).
|
|