Core Messaging deck “Data Access – A New Approach”
Transcripción
Core Messaging deck “Data Access – A New Approach”
Data Access—A New Approach October 11, 2016 10/11/2016 © Gen-Z Consortium 2016 1 Gen-Z Consortium • Industry Leaders developing a memory-semantic interconnect 10/11/2016 © Gen-Z Consortium 2016 2 Why Propose a New Bus? Explosive Growth of Data • System memory is flat or shrinking • Memory bandwidth per core continues to decrease • Memory capacity per core is generally flat • Memory is changing on a different cadence compared to the CPU 2015 2005 2010 0.1ZB 1.2ZB 2.8ZB 8.5ZB • Data is growing • More than 37% of total data generated in 2020 (40 ZB) will have Big Data value • Data that requires real-time analysis is growing exponentially • The value of the analysis decreases if it takes too long to provide insights Need Answers … FAST! • The industry needs an open architecture to solve the problems 10-2 100 102 104 106 Seconds Time to Result • • © Gen-Z Consortium 2016 Value of Analyzed Data $ • Memory tiers will become increasingly important • Rack-scale composability requires a high bandwidth, low latency fabric • Must seamlessly plug into existing ecosystems without requiring OS changes 10/11/2016 2020 40 ZB 2012 Businesses demanding real-time insight Increasing amounts of data to be analyzed 3 Memory/Storage Convergence: The Media Revolution Today Memory DRAM DRAM SCM* Storage Disk/SSD Disk/SSD Memory Semantics will be pervasive in Volatile AND NonVolatile Storage as these technologies continue to converge. *SCM = Storage Class Memory **OPM = On-Package Memory 10/11/2016 © Gen-Z Consortium 2016 DRAM/OPM** SCM* DRAM/OPM** SCM* Disk/SSD Disk/SSD New and Emerging Memory Technologies HMC 3DXPointTM Memory HBM MRAM Low Latency NAND PCM Managed DRAM RRAM 4 Memory-Semantic Fabric • Communication at the speed of Memory What is a Memory Semantic Fabric? • Handles all communication as memory operations such as load/store, put/get and atomic operations typically used by a processor Compute SoC SoC Memory Memory Accelerators FPGA FPGA GPU GPU Memory Memory • Memory semantics are optimal at sub-microsecond latencies from CPU load command to register store • Unlike, storage accesses which are block based and managed by complex, code intensive, software stacks Memory Semantics Why Now? • The emergence of low latency, Storage Class Memory (SCM) and the demand for large capacity, rack scale resource pools, and multi node architectures 10/11/2016 © Gen-Z Consortium 2016 Pooled Memory Network Storage I/O 5 Gen-Z: A New Data Access Technology High Bandwidth Low Latency Advanced Workloads & Technologies − Memory Semantics – simple Reads and Writes − From tens to several hundred GB/s of bandwidth Compute − Sub-100 ns load-to-use memory latency SoC SoC − Real time analytics Memory Memory Accelerators FPGA FPGA GPU GPU Memory Memory − Enables data centric and hybrid computing − Scalable memory pools for in memory applications − Abstracts media interface from SoC to unlock new media innovation Gen-Z Direct Attach, Switched, or Fabric Topology − Provides end-to-end secure connectivity from node level to rack scale Secure Compatible Economical − Supports unmodified OS for SW compatibility Pooled Memory − Graduated implementation from simple, low cost to highly capable and robust − Leverages high-volume IEEE physical layers and broad, deep industry ecosystem 10/11/2016 © Gen-Z Consortium 2016 Network Storage I/O Rack Scale 6 Gen-Z Consortium Mission • Create a next generation interconnect that will bridge existing solutions while enabling new unbounded innovation • Develop in an open, non-proprietary standards body where adoption, differentiation and innovation is promoted as an industry standard. Board Members A Transparent Organization: Gen-Z has been formed as a not-for-profit organization, its ongoing development occurs on the basis of an open decision-making procedure available to all interested parties. Wide availability: The Gen-Z standard will be published and available free of charge. End-User Choice: There are no constraints on the re-use of the standard. Gen-Z creates a fair, competitive market for implementations of the standard. Equality: Gen-Z does not favor one implementer over another. 10/11/2016 © Gen-Z Consortium 2016 7