Memory At Storage Scale Storage At Memory Speed
Memory at Storage ScaleStorage at Memory SpeedCharles Fan, Co-founder and CEO, MemVerge1
Big Data and Fast DataStorage IO is slow(100’s of microseconds)DRAM is small andexpensive(100’s of GBs)Could there be a solution that makes storage faster and memory bigger?Machine Learning, Big Data and IoT demandnanosecond speed petabyte scale datainfrastructure 2020 SNIA Persistent Memory Summit. All Rights Reserved2
Persistent Memory Intel Delivered Optane DC Persistent Memorybased on 3D XPoint tech in Q2 2019.– Revenue projected to reach 3.6B by 2023/2024– Additional major vendors to join the foray by 2022 Storage at Memory Speed– 100-250ns R/W latency Memory at Storage Scale– 6TB/server– Persistent3 2020 SNIA Persistent Memory Summit. All Rights Reserved
World’s First Memory Converged InfrastructureMemVerge software leverages Storage Class Memorytechnology to deliver larger memory and faster storage toapplications without requiring application rewritesPMEM-native DistributedFile SystemPMEM-native DistributedMemory SystemMemory “Hypervisor”4 2020 SNIA Persistent Memory Summit. All Rights ReservedMemVerge Confidential4
Architecture and Use CasesMachineLearningSearch and QueriesBig Data AnalyticsTrading / Market Data.Pub/SubDistributed Memory Objects (DMO)PMEM-native Distributed FileSystemPMEM-native DistributedMemory SystemMemory “Hypervisor”Compute Node 1DRAMPMEMSSD 2020 SNIA Persistent Memory Summit. All Rights ReservedCompute Node 2DRAMPMEMSSDCompute Node 3DRAMPMEMSSDCompute Node 4DRAMPMEMSSD.Compute Node 5DRAMPMEMSSD5
Use Case: Big Data Analytics with Spark Problem Spark SQL Out of DRAM Disk I/O too slow Data spill degrades performance Local SSDs wear out by frequent intermediate data writes Solution Adding MemVerge DMO to the Spark cluster accelerates theentire cluster Moving intermediate state off Spark Elastic Computingnodes increased the cloud elasticity of the solution5X7X100%Terasort SpeedRDD Caching SpeedCloud Elasticity66 2020 SNIA Persistent Memory Summit. All Rights Reserved
Use Case: AI Training with Checkpointing Problem Model training takes a long time to complete for large datasets Failure recovery is painful without frequent checkpointing Data preprocessing and importing can take a long time Delayed model deployment Solution MemVerge DMO, powered by Optane DC persistent memory,improves checkpointing speed and data loading speedup to6XTraining Speedup to350XData Import SpeedInstantCheckpoint Recovery77 2020 SNIA Persistent Memory Summit. All Rights Reserved
Use Case: High Performance Financial Systems High Performance Market Data Service– Enables high bandwidth and low latency marketdata replay via large memory-mapped files Pub/Sub Service– Next-gen Pub/Sub architecture with RDMAinterconnected Persistent Memory KDB Acceleration– Improve the performance of existingdeployments of KDB 2020 SNIA Persistent Memory Summit. All Rights Reserved88
MemVerge Vision By 2025, Persistent Memory will be mainstream. Data Infrastructure will be memorycentric. Performance-tier storage will be replaced by Memory Converged Infrastructure (MCI)co-located with compute. Memory capacity will be expanded by the same MCI layer. MemVerge aspires to be a leader in MCI.ComputeMemoryComputeMCIPerformance-tier StorageCapacity-tier StorageTodayCapacity-tier StorageTomorrow9 2020 SNIA Persistent Memory Summit. All Rights Reserved
Thank youPlease visit www.snia.org/pmsummit for presentations10
Memory at Storage Scale Storage at Memory Speed Charles Fan, Co-founder and CEO, MemVerge. 2 DRAM is small and expensive (100’s of GBs) Storage IO is slow (100’s of microseconds) Could there be a solution that makes storage faster and memory
Memory Management Ideally programmers want memory that is o large o fast o non volatile o and cheap Memory hierarchy o small amount of fast, expensive memory -cache o some medium-speed, medium price main memory o gigabytes of slow, cheap disk storage Memory management tasks o Allocate and de-allocate memory for processes o Keep track of used memory and by whom
May 01, 2018 · 7. Three differences between memory and storage: Memory is faster than storage. Memory is volatile (it only keeps its contents whilst it is supplied with power) and storage is not. Memory is more expensive per GB than storage. Memory is electronic whereas storage can b
CCC-466/SCALE 3 in 1985 CCC-725/SCALE 5 in 2004 CCC-545/SCALE 4.0 in 1990 CCC-732/SCALE 5.1 in 2006 SCALE 4.1 in 1992 CCC-750/SCALE 6.0 in 2009 SCALE 4.2 in 1994 CCC-785/SCALE 6.1 in 2011 SCALE 4.3 in 1995 CCC-834/SCALE 6.2 in 2016 The SCALE team is thankful for 40 years of sustaining support from NRC
In memory of Paul Laliberte In memory of Raymond Proulx In memory of Robert G. Jones In memory of Jim Walsh In memory of Jay Kronan In memory of Beth Ann Findlen In memory of Richard L. Small, Jr. In memory of Amalia Phillips In honor of Volunteers (9) In honor of Andrew Dowgiert In memory of
21-07-2017 2 Chap. 12 Memory Organization Memory Organization 12-5 12-1 Memory Hierarchy Memory hierarchy in a computer system Main Memory: memory unit that communicates directly with the CPU (RAM) Auxiliary Memory: device that provide backup storage (Disk Drives) Cache Memory: special very-high-
An Introduction to Memory LO 1 Define memory. LO 2 Describe the processes of encoding, storage, and retrieval. Flow With It: Stages of Memory LO 3 Explain the stages of memory described by the information-processing model. LO 4 Describe sensory memory. LO 5 Summarize short-term memory. LO 6 Give examples of how we can use chunking to improve our memory span.
Chapter 2 Memory Hierarchy Design 2 Introduction Goal: unlimited amount of memory with low latency Fast memory technology is more expensive per bit than slower memory –Use principle of locality (spatial and temporal) Solution: organize memory system into a hierarchy –Entire addressable memory space available in largest, slowest memory –Incrementally smaller and faster memories, each .
Events notification (hooks) in real time Webhooks are calls made to your custom URL when any event gets fired. You can define your own hooks URL at client and account levels.
