Upgrade to Pro — share decks privately, control downloads, hide ads and more …

VAST Data - IT Press Tour June 2020

VAST Data - IT Press Tour June 2020

Avatar for The IT Press Tour

The IT Press Tour

June 12, 2020
Tweet

More Decks by The IT Press Tour

Other Decks in Technology

Transcript

  1. C O N F I D E N T I

    A L I N T R O D U C T I O N
  2. 2 B R E A K T H R O

    U G H INFRASTRUCTURE CONCEPT V A S T D A T A COMPANY INTRODUCTION P I O N E E R I N G CUSTOMER STORIES C O N F I D E N T I A L O V E R V I E W
  3. 3 B R E A K T H R O

    U G H INFRASTRUCTURE CONCEPT V A S T D A T A COMPANY INTRODUCTION P I O N E E R I N G CUSTOMER STORIES C O N F I D E N T I A L O V E R V I E W
  4. CORPORATE TIMELINE 4 2016 2017 2018 2019 2020 Company Founded

    20+ End User Trials v1.0 GA v2.0 GA v3.0 GA Series A = $15M Exits Stealth- Mode Series B = $40M Series A1 = $25M Record- Breaking Y1 Series C = $100M $1.2B Valuation C O N F I D E N T I A L O V E R V I E W
  5. A PROVEN MANAGEMENT TEAM 5 Renen Hallak CEO and Founder

    Mike Wing President Shachar Feinblit VP, R&D and Co-Founder Avery Pham VP, Operations Jeff Denworth VP, Products & Co-Founder C O N F I D E N T I A L O V E R V I E W
  6. z Record-Shattering Year-1 Business Performance AN UNPARRALELED EARLY SUCCESS STORY

    6 First Full Year Performance Gross Profit Revenue z z z Customer Adoption Dozens Across 4 Continents Average Customer Y1 Spend $1,020,000 Selling Like a Unicorn, Spending Like a Camel Clear Path to Breakeven; $140M of Cash in Bank C O N F I D E N T I A L O V E R V I E W
  7. 7 B R E A K T H R O

    U G H INFRASTRUCTURE CONCEPT V A S T D A T A COMPANY INTRODUCTION P I O N E E R I N G CUSTOMER STORIES C O N F I D E N T I A L O V E R V I E W
  8. 30 YEARS OF STORAGE COMPLEXITY 8 All-Flash Arrays RAM +

    3D XPoint Backup NAS Archive Object, Cloud, Tape C O N F I D E N T I A L O V E R V I E W
  9. 9 ON-PREMISES INFRASTRUCTURE MUST BE AS SIMPLE AS PUBLIC CLOUD

    C O N F I D E N T I A L O V E R V I E W
  10. 1 0 NEW MACHINE LEARNING WORKLOADS DEMAND RANDOM, WIDE ACCESS

    TO DATA C O N F I D E N T I A L O V E R V I E W
  11. 1 1 PERFORMANCE CAPACITY OUR FOUNDATIONAL REALIZATION: FLASH BREAKS THE

    PERFORMANCE vs. CAPACITY TRADOFF C O N F I D E N T I A L O V E R V I E W
  12. 1 2 Welcome to the Universal Storage Era INTRODUCING VAST

    DATA Write at 3D XPoint speeds, Read at TB/s, Ms of IOPS ALL-NVMe PERFORMANCE exceeds the capacity needs of any size organization EXABYTE-SCALE FILE & OBJECT STORAGE Engineered at every level to deliver unrivaled system efficiency TIER 5 COST EFFICIENCY C O N F I D E N T I A L O V E R V I E W
  13. 1 3 Technologies not available to storage companies before 2018

    A FOUNDATION FOR A NEW ARCHITECTURE NVME OVER FABRICS 3D XPoint Intel Optane Memory LOW-COS FLASH C O N F I D E N T I A L O V E R V I E W
  14. Powered By VAST’s Disaggregated, Shared Everything (DASE) Architecture INTRODUCING: VAST

    DATA UNIVERSAL STORAGE 3D XPoint QLC Flash NVMe Enclosure N V M E F A B R I C : C O M M O D I T Y E T H E R N E T O R I N F I N I B A N D CLIENTS: NFS, NFSoRDMA, SMB, S3, K8S CSI VAST Storage Servers (Containers) 3D XPoint QLC Flash NVMe Enclosure 3D XPoint QLC Flash NVMe Enclosure 3D XPoint QLC Flash NVMe Enclosure G L O B A L N A M E S P A C E 1 4 C O N F I D E N T I A L O V E R V I E W
  15. DASE ARCHITECTURE: ’GLOBAL’ ADVANTAGES 3D XPoint QLC Flash NVMe Enclosure

    1 0 µ s D I S T A N C E T O S T O R A G E S Y S T E M S T A T E 3D XPoint QLC Flash NVMe Enclosure 3D XPoint QLC Flash NVMe Enclosure 3D XPoint QLC Flash NVMe Enclosure G L O B A L N A M E S P A C E 1 5 A FOUNDATION FOR GLOBAL ALGORITHMS global namespace | global flash translation | global data protection | global data reduction STATELESS ARCHITECTURE, ELIMINATES EAST-WEST TRAFFIC no cache coherency challenges | no batteries | no rebuilds during server failure | docker-based auto-scaling C O N F I D E N T I A L O V E R V I E W
  16. VAST SERVER POOLS ENABLE MULTI-TENANCY 3D XPoint QLC Flash NVMe

    Enclosure 3D XPoint QLC Flash NVMe Enclosure 3D XPoint QLC Flash NVMe Enclosure 3D XPoint QLC Flash NVMe Enclosure G L O B A L N A M E S P A C E 1 6 One Storage System To Consolidate Noisy Neighbors & Islands of Storage S E G M E N T B Y A P P L I C A T I O N O R N E T W O R K S E G M E N T B Y A P P L I C A T I O N O R N E T W O R K C O N F I D E N T I A L O V E R V I E W
  17. DEPLOYMENT OPTIONS 1 7 ENCLOSURE + SERVER APPLIANCE ENCLOSURE +

    VAST CONTAINERS SOFTWARE-ONLY SSD SSD SSD SSD SCM SCM C O N F I D E N T I A L O V E R V I E W
  18. DEPLOYMENT OPTIONS: HW APPLIANCE 1 8 VAST HA NVMe Enclosures

    400TB and 675TB Options 4 x 100Gb NVMeF Connectivity 2U of Space NVMe Fabric 100Gb Ethernet or InfiniBand VAST Container Servers 4 Servers in 2U Chassis Ethernet and InfiniBand Connectivity Capacity-Optimized vs. HDD Options: - Scale-Out NAS: 6.7PB - VAST Data @2:1: 23PB (2.4x more dense) Performance-Optimized vs. Flash Options: - Scale-Out NAS: 2.3PB - VAST Data @2:1: 11PB (3.8x more dense) C O N F I D E N T I A L O V E R V I E W
  19. DATA IS NEVER CONTAINED OR BOTTLENECKED BY A GATEWAY MULTI-PROTOCOL

    NAMESPACE 1 9 SMB S3 NFS tbd Protocol Layer Attributes: Rich, Shared-Everything Metadata Structures POSIX & S3 metadata, multi-protocol mappings, links, lock state, snap state, keys Data store: Capacity-Efficient a byte-granular, thin-provisioned, sharded, infinitely-scalable data store Element Store C O N F I D E N T I A L O V E R V I E W
  20. C O N F I D E N T I

    A L O V E R V I E W 2 0 L E G A C Y S H A R E D - N O T H I N G S T O R A G E THE PROBLEM WITH STATEFUL PROTOCOLS legacy NAS systems do not effectively mirror SMB state across all controllers locks, leases in DRAM
  21. C O N F I D E N T I

    A L O V E R V I E W 2 1 L E G A C Y S H A R E D - N O T H I N G S T O R A G E THE PROBLEM WITH STATEFUL PROTOCOLS “with isilon, because it takes so long to do the upgrades in a rolling manner it is just hard to say for the next two days your smb connection may drop at any time” - customer quote from product requirements gathering session
  22. QLC Flash HA Enclosure QLC Flash HA Enclosure QLC Flash

    HA Enclosure QLC Flash HA Enclosure C O N F I D E N T I A L O V E R V I E W 2 2 N V M e F a b r i c G L O B A L N A M E S P A C E INTRODUCING: RESILIENT SMB stateless 3D XPoint 3D XPoint 3D XPoint 3D XPoint metadata
  23. C O N F I D E N T I

    A L O V E R V I E W 2 3 3D XPoint QLC Flash HA Enclosure N V M e F a b r i c 3D XPoint QLC Flash HA Enclosure 3D XPoint QLC Flash HA Enclosure 3D XPoint QLC Flash HA Enclosure G L O B A L N A M E S P A C E INTRODUCING: RESILIENT SMB State, locks, leases: fail over in microseconds. Upgrades are simple. Uptime is mastered.
  24. NOW WIITH CLOUD BACKUP C O N F I D

    E N T I A L O V E R V I E W 2 4 all flash on prem cloud
  25. Global QLC Flash Translation New Global Storage Algorithms Break Decades

    of Tradeoffs GAME CHANGING STORAGE INNOVATIONS 2 5 Global Erasure Codes Global Data Reduction QLC Flash Saves 85% Save Up To Another 66% Unprecedented Efficiency MLC/ TLC QLC COST C O N F I D E N T I A L O V E R V I E W
  26. Write Amplification Block Size = 4KB Flash Page Size =

    64KB Erase Block = 200MB #1 4K Write Consumes 64K Page #2 Garbage Collection #3 Short lived data ages out #4 More Garbage Collection +16x +2x +2x… C O N F I D E N T I A L O V E R V I E W 2 6
  27. 2 7 C O N T R O L L

    E R C O N T R O L L E R C O N T R O L L E R C O N T R O L L E R DI ST RI B UT E D, P E RSI ST E NT XP OI NT B UFFE R 100x LARGER THAN DRAM C O N F I D E N T I A L O V E R V I E W
  28. 2x more longevity than HDD-based systems BREAKING THE PRICE/ENDURANCE TRADEOFF

    10–YEAR ENDURANCE WARRANTY C O N F I D E N T I A L O V E R V I E W 2 8
  29. 15x-30x more efficient than competing approaches GLOBAL DATA PROTECTION C

    O N F I D E N T I A L O V E R V I E W 2 9 F L A S H B U F F E R X P O I N T X P O I N T X P O I N T X P O I N T • Shared Nothing Designs • Small, Volatile Write Caches • Reed-Solomon Historic Barriers to Wide Striping VAST codes accelerate rebuild speed by using a new type of algorithm that gets faster with more redundancy data. • 36+4: Much faster than classic RAID, more resilience, 9% overhead • 146+4: 60M years of mean-time-to-data-loss, only 2.6% overhead VAST’s Next-Generation Locally-Decodable Erasure Encoding
  30. A breakthrough in data reduction SIMILARITY-BASED, GLOBAL DATA REDUCTION C

    O N F I D E N T I A L O V E R V I E W 3 0 Data is fingerprinted in large blocks after the write is persisted in SCM Fingerprints are compared to measure relative distance, similar chunks are clustered Clustered data is compressed together; byte-level deltas are extracted & stored DELTAS REFERENCE subsequent reads are serviced within 1ms using locally decodable compression algorithms
  31. DATA REDUCTION IN PRACTICE 3 1 LIFE SCIENCE 2:1 ANIMATION

    3:1 SEARCH 4:1 BACKUP APPLIANCES 3:1 pre-compressed and pre-deduplicated pre-compressed pre-compressed pre-compressed un-compressed HPC 3:1 MARKET DATA 8:1 pre-compressed BACKUP 20:1 C O N F I D E N T I A L O V E R V I E W
  32. 3 2 NAS All Flash Cloud, Object, Archive Universal Storage

    THEN NOW Simplify the stack. Unleash insights. C O N F I D E N T I A L O V E R V I E W
  33. 3 4 VS. flash tiering needs endurance Legacy Storage Thinking

    endurance is amortized C O N F I D E N T I A L O V E R V I E W
  34. 3 5 VS. NFS is an old file system Legacy

    Storage Thinking NFS is merely a means of transport C O N F I D E N T I A L O V E R V I E W
  35. UNIVERSAL STORAGE. FAST, SIMPLE, AFFORDABLE. C O N F I

    D E N T I A L O V E R V I E W 3 6 Up to 200GB/s per Client (100x Faster Than. TCP NFS) Linear Scaling to Exabytes NAS (over RDMA) Simplicity 80% Less Cost N O T R A D E O F F S THE NEW AI STORAGE PARADIGM T O D A Y ’ S S H A D O W A I O P T I O N S SCALE-OUT NAS Simple & Versatile, But Slow HPC FILE SYSTEMS Fast & Scalable, But Complex Both Are Prohibitively Expensive for All-Flash
  36. C O N F I D E N T I

    A L O V E R V I E W 3 7 DISAGGREGATION IS THE PATH TO EMBARASSINGLY PARALLEL SCALE ONLY VAST IS TRULY LINEAR N V M e F A B R I C G L O B A L N A M E S P A C E
  37. LEGACY FLASH ARCHITECTURES TIERED STORAGE ARCHITECTURES VAST DATA UNIVERSAL STORAGE

    EFFECTIVE COST/PB HDD Flash C O N F I D E N T I A L O V E R V I E W 3 8 PIONEERING RADICAL FLASH SAVINGS VAST ALL-FLASH IS IDEAL FOR LAUNCHING NEXT-GEN AI INITIATIVES $2M $760K $400K Universal Storage brings flash TCO in line with tiered storage approaches to de-risk customer storage decisions all while freeing up HW capital for GPUs. 200TB @ $3.5/GB 800TB @ $0.2/GB QLC + 2.5% Erasure Codes + Similarity-Based Data Reduction Our formula for compounded flash savings: C O N F I D E N T I A L O V E R V I E W 3 8
  38. VAST USE CASES 3 9 V A S T U

    N I V E R S A L S T O R A G E : O N E P L A T F O R M F O R A L L D A T A C E N T E R D A T A Secondary Storage Big Data & AI Vertical Solutions Genomics Finance Content HPC Enterprise Infrastructure C O N F I D E N T I A L O V E R V I E W
  39. 4 0 B R E A K T H R

    O U G H INFRASTRUCTURE CONCEPT V A S T D A T A COMPANY INTRODUCTION P I O N E E R I N G CUSTOMER STORIES C O N F I D E N T I A L O V E R V I E W
  40. UNIVERSAL STORAGE FOR THE WORLD’S INFORMATION C O N F

    I D E N T I A L O V E R V I E W 4 1 A I • Q U A N T I T A T I V E T R A D I N G • L I F E S C I E N C E S • M E D I A • S E A R C H A N I M A T I O N & V F X • B A C K U P • C L U S T E R C O M P U T I N G • C O N T A I N E R S
  41. CUSTOMER QUOTES 4 2 “As our component Operating Divisions move

    beyond the hard drive era, software-enabled storage architectures helps us modernize our scientific agenda and enable AI- driven research with the power of flash.” Jose Arrieta, CIO of HHS “We invest in many cutting-edge technologies, which includes VAST Data’s Universal Storage platform, to support our most intensive computing efforts, for hundreds of researchers on a global scale" Olivier Delahaye, CTOof Squarepoint “VAST provides Zebra a solution to all of our A.I. storage challenges by delivering performance superior to what is possible with traditional NAS while also providing a simple, scalable appliance that requires no effort to deploy and manage.” Eyal Toledano, CTO of Zebra Medical ““VAST Data provides Ginkgo the potential to ride the declining cost curve of flash while also providing near- infinite scale.” Austin Che, CTO of Ginkgo Bioworks C O N F I D E N T I A L O V E R V I E W
  42. Linear Performance at Scale 4 3 CBOX x DBOX Capacity

    (TBu @ 2:1) Random Writes (GB/s) Random Read 2019 Enclosure (GB/s) Random Read 2020 Enclosure (GB/s) IOPS (4K random Read) 1x1 1.1PBu 5 20 40 225K 2x2 2.3PBu 10 40 80 500K 3x3 3.5PBu 15 60 120 750K 4x4 4.6PBu 20 80 160 1M 5x5 5.7PBu 25 100 200 1.25M 6x6 6.8PBu 30 120 240 1.5M 7x7 7.9PBu 35 140 280 1.75M 8x8 9.1PBu 40 160 320 2M 9x9 10.3PBu 45 180 360 2.25M At 20GB/s per enclosure, NIH saw linear scale to 9 of VAST’s 2019 enclosures in a single cluster C O N F I D E N T I A L O V E R V I E W
  43. C O N F I D E N T I

    A L O V E R V I E W 4 4 18TB 3D XPoint 675TB QLC Flash HA Enclosure Render Servers 7x faster than Isilon Character Cache Servers 2ms or less, Same as Avere AI GPU Machines Unrivaled NAS Throughput – 5x Faster than Isilon
  44. C O N F I D E N T I

    A L O V E R V I E W 4 5 ingest editing enterprise Universal Storage in Action 18TB 3D XPoint 675TB QLC Flash HA Enclosure 18TB 3D XPoint 675TB QLC Flash HA Enclosure 18TB 3D XPoint 675TB QLC Flash HA Enclosure
  45. C O N F I D E N T I

    A L O V E R V I E W 4 6 I N I T I A L O R D E R 10 PB ($Ms) A D D I T I O N A L U P S I D E 2,000+ PB (SW) U N S E A T I N G NetApp in one their largest accounts V A S T A D V A N T A G E DAS Economics. Appliance Simplicity. Exabyte Scale. C A S E S T U D Y LEADING US TELCO