Paper Title: Deep Dive on Disaggregated Storage Architectures: Challenges and Solutions

Paper Abstract: Low latency flash is becoming an increasingly popular choice for enabling latency sensitive workloads. It is not uncommon to expect these devices providing million IOPS. Deploying high performance devices on compute nodes discourages sharing, resulting in increased TCO. It is economical to deploy pool of high performance flash in a rack and provision storage at run-time for ephemeral and durable storage use cases. However, carving out storage from the flash pool and exposing over network introduces latency. In this session we outline different network protocols, latency implication and industry trends to deliver low latency dis-aggregated flash storage over the network.

Paper Author: Vaidyanathan Krishnamoorthy, Cloud Solutions Architect, Intel
Mrittika Ganguli, Network Architect, Intel

Author Bio: Vaidy is a Cloud Solutions Architect in Intel Data Center Solutions Group. He is responsible for solutions involving Software Defined Storage architecture and technology initiatives within Intel. He has over 22 years of Software development across Mobile Phone Platforms, Connected TV, Internet-of-Things End-to-End solutions and Cloud Solutions involving Network Virtualization and Software Defined Storage.

Author 2 Bio: Mrittika Ganguli is a Principle Engineer and Network Software Architect in Intel DataCenter Group, with 20+ years of experience in Cloud, Network, storage, open source orchestration software and Systems Design, and Architecture. Her recent work at Intel lies defining the network and telemetry architecture specifications for Intel Rackscale Design (RSD) deployed for cloud in datacenters. She has more than 30 patents published.