Presentation for the registation gONI.pptx

office of naval int

1. Agenda Technology Discussions Data Life Cycle and Analytics Pure Offerings AI/Data Analytic Infrastructure(s) Q&A

2. Technology Discussion Areas Data Life Cycle Shifts (today’s discussion) Datafication of everything Storage, access, and TCO Increasing the value of data through Anaylitcs Infrastructure as Code Virtual Data Center Ubiquitous Multi-Premise Mobility (Hybrid Cloud) Digital Transformation Access to Everything from Anywhere Secure End-User Mobility

3. Mega Trend - Datafication of Everything

4. Data Analytic Trends Requirement for learning on data sets growing larger Analytics is driving a change in the Data Life Cycle Architecture parallelization from compute to storage ( NVMe ) to meet performance demands (AI Optimized Infrastructures) Distributed Multi-Premise (hybrid) model learning – processing close to the data at disparate locations Heterogeneous machine learning/AI architectures with data sharing End users interacting directly with the technology without need for programming or data scientist expertise ML/AI Visualization On-Premise Cloud Flash Object

5. Shifting the Data Life Cycle OLD ERA Offsite Tape Facility Data Value = 0 Insurance Policy for Data Cold Data as Archive MODERN ERA Cloud Economics & Durability Reuse Data Cold Data as Value-Creating Asset

6. Pure Technology Family FlashArray All Flash block level storage array Evergreen Storage changes TCO model FlashBlade All Flash file level storage array Massively Parallel scale out performance supports HPC, AI, and Data Analytic Object Engine Designed to enable hybrid cloud data strategies Data reduction and optimization engine minimized transnational and storage costs Hardware Appliance (OE/FB) delivered on- prem Virtual Appliance runs in the cloud *

7. Disrupting the Market

8. Notional Architecture RAPID RECOVERY PRIMARY FLASH DEDUPE FLASH STORE for Most Recent Data PBBA PBBA PRIMARY DISK DISK BACKUP APPLIANCE BACKUP DISK BACKUP APPLIANCE TAPE SLOW RECOVERY PBBA ARRAY PBBA PBBA PBBA TAPE LIB LEGACY: D2D2T MODERN: F2F2C Production Site Offsite Natively Replicated All Data 11 9’s of Durability MEET SLAs w/ Rapid Restore SAVE MONEY w/ Cloud Economics REUSE DATA W/ Cloud Apps

9. 1PB in 3U for ~1500 Watts MULTI- PROTOCOL BUILT-IN 1PB IN JUST 3U 100% NVMe DENSITY: 12 GB/s BANDWIDTH 100% NVMe PERFORMANCE: 99.9999% AVAILABILITY & NDU EVERYTHING NOW 100% NVMe

10. Evergreen - Saves 61% Over 6 Years

11. FlashBlade

12. FlashBlade

13. FlashBlade References

14. BORN FROM FLASH + CLOUD OE//A270 4-Nodes 25 TB/HR Backup 15 TB/HR Restore 15 PB Front-end Capacity Inline Dedup to Flash & Cloud System GA: Q1 2019 CloudDirect GA: Q2 2019 OE//Cloud Scales to 100+TB/HR Backup Protects 100+ PB in Cloud 11 9’s of Durability, Internally Replicated Global Namespace, S3API GA: 2H2019

15. AIRI - AI Ready Infrastructure +

16. AIRI Scaling Matrix FLASH BLADE Chassis 1 2 3 4 5 DGX-1 Systems 1 Compute: 1 PFLOP Throughput: 17 GB/Sec Storage: 1.6 PB Space: 8U 1 PFLOP 34 GB/Sec 3.2 PB 14U 1 PFLOP 51 GB/Sec 4.8 PB 18U 1 PFLOP 68 GB/Sec 6.4 PB 22U 1 PFLOP 85 GB/Sec 8.0 PB 26U 2 2 PFLOPs 17 GB/Sec 1.6 PB 11U 2 PFLOPs 34 GB/Sec 3.2 PB 17U 2 PFLOPs 51 GB/Sec 4.8 PB 21U 2 PFLOPs 68 GB/Sec 6.4 PB 25U 2 PFLOPs 85 GB/Sec 8.0 PB 29U 3 3 PFLOPs 17 GB/Sec 1.6 PB 14U 3 PFLOPs 34 GB/Sec 3.2 PB 20U 3 PFLOPs 5 1GB/Sec 4.8 PB 24U 3 PFLOPs 68 GB/Sec 6.4 PB 28U 3PFLOPs 85 GB/Sec 8.0 PB 32U 4 4 PFLOPs 17 GB/Sec 1.6 PB 17U 4 PFLOPs 34 GB/Sec 3.2 PB 23U 4 PFLOPs 51 GB/Sec 4.8 PB 27U 4 PFLOPs 68 GB/Sec 6.4 PB 31U 4PFLOPs 85 GB/Sec 8.0 PB 35U

17. IBM – Meet Summit Summit Performance Peak Performance = 200 petaflops 10 PB of memory Summit Configuration 4,608 IBM AC922 servers each with two 22 core IBM Power9 processors 202,752 total cores. Each server also has 6 Nvidia Tesla V100 GPUs 27,648 Nvidia v100 GPUs 17,694,720 Tensor Cores File system IO >2TB per second Power Consumption 13 megawatts 2021 – ORNL expects to take delivery of Frontier, the first ExaFlop Supercomputer

18. Accelerated Compute Platform (ACP) - Mini-CORAL Summit Reference AI Optimized Architecture Summit Performance at Scale

19. Q & A