FLASH.NEXT. Zero to One Million IOPS in a Flash. Ahmed Iraqi Account Systems Engineer North Africa

1 FLASH.NEXT Zero to One Million IOPS in a Flash Ahmed Iraqi Account Systems Engineer North Africa

2 The Performance Gap CPU performance increases 100x every decade Hard disk drive performance has stagnated Xeon E7-4800 Pentium Pro I/O Gap 1995 2012 Hard Disk Drive IOPS/Spindle

3 Expectations Are Reset Forever

4 DATA IS GROWING

5 While At The Same Time Costs Must Be Contained Information Must Become An Asset Performance Must Be Delivered At Massive Scale

6 ONLY FLASH CAN MEET THESE EXPECTATIONS

7 Flash: Redefining The Possible 2000 IOPS/GB 4000X FASTER THAN HARD DRIVES 0.5 IOPS/GB Hard Disk Drive PCIe Flash

8 HOW DO WE BEST DEPLOY FLASH?

9 XtremSF Performance Efficiency Flexibility

10 Introducing EMC XtremSF Flexibility Broad Range Of emlc And SLC Capacities For Use As Local Storage Or With VFCache Performance Dramatic Improvements In Application Latency And Throughput Up To 1.13 Million IOPS Efficiency Best TCO Lowest Overhead, Highest Performance, Lowest Power Consumption, Best Density

11 The XtremSF Family All Cards Are HHHL Highest Density In The Industry XtremSF 2200 2.2 TB XtremSF 1400 1.4 TB XtremSF 700 700 GB XtremSF 550 550 GB UP TO 1,130,000 IOPS

13 EMC XtremSW Cache Server Flash DAS Becomes Cache Performance Dramatic Improvements In Latency And Throughput Intelligence Extends EMC FAST Architecture Into The Server Protection Backed By Intelligent, Resilient, Highly Available Storage

Latency Latency TPS TPS 14 EMC XtremSW Cache Application Testing Results 7.7X FASTER RESPONSE TIME 1.0 3.6 3.6X BETTER PERFORMANCE 0.13 1.0 2.5X FASTER RESPONSE TIME 1.0 3.1 3.1X BETTER PERFORMANCE 0.4 1.0

15 XtremSW Cache Split Card Mode The Best Of Both Worlds Cache Direct- Attached Storage Data 1 Data 2 Temp DB

Working Set Copyright 2013 EMC Corporation. All rights reserved. 16 Use Cases Small Temp tables External sorts Working files Write-intensive workloads with small working sets Write-intensive workloads with large working sets CRM ERP OLTP Email Web Servers Reporting Analytics Read-intensive workloads with small working sets Read-intensive workloads with large working sets Large Mostly Writes Backups Write journals Oil & gas Data dumps Image capture Workload Big Data Streaming Genetics studies Mostly Reads

17 Traditional Architecture Server Network Storage Application CPU Switch Storage Array SSD/HDD 1 2 3 4 Flash FC / SAS 8 7 6 5 SATA / NL-SAS Reads (latency: 500 μs 10 ms) Writes Reads and writes are serviced by the storage array Performance varies depending on back-end array s media, workload, and network

18 XtremSW Cache Advanced Architecture Server Network Storage Application CPU PCIe Flash Card Switch Storage Array SSD/HDD 1 2 3 4 Flash 5 FC / SAS 9 8 7 6 SATA / NL-SAS Reads (latency: <100 μs) Writes Reads are serviced by XtremSW Cache for Writes are passed through to the storage array for

19 Read Hit Example SAN HBA 1 Application XtremSW Cache Driver 2 SAN storage 3 XtremSF 1. Read request from application to an accelerated array LUN 2. XtremSW Cache Driver determines a hit occurred and accesses data from Flash device 3. Data returned from the Flash device is forwarded to the application

20 Read Miss Example Application 1 1. Read request from application to an accelerated array LUN 2 XtremSW Cache Driver 4 SAN HBA 3 XtremSF SAN storage 2. XtremSW Cache Driver determines a miss occurred and accesses data from array LUN 3. Data is read from the array and returned to the application 4. Read miss data is written to Flash device

21 Write Example 1 Application 1. Write request from application to an accelerated array LUN 2 XtremSW Cache Driver 2 SAN HBA 3 XtremSF SAN storage 2. XtremSW Cache driver writes data to array LUN and Flash device 3. Application write acknowledged upon array completion

22 WHAT S COMING? EMC Array Integration Distributed Cache Coherency Advanced VMware Support Additional Environments AIX, SLES, 3 rd Party PCIe

Array Integration VMAX Single Point Of Management LUN Selection Based On VMAX Trending Analysis Call-home Support Integrated Performance Statistics Reporting VNX Single Point Of Management Performance And Health Monitoring Discovery And Configuration LUN Selection Based On VNX Trending Analysis Unmatched Performance Boost Cache Coordination: Optimized Read Miss Prefetching Of Data To XtremSW Cache Copyright 2013 EMC Corporation. All rights reserved. 23

24 XtremSW Cache Future Direction Cache Coherency For Oracle RAC, Broad PCIe & OS Support PCIe Card Support Additional OS Support

25 XtremSW Suite Future Direction Data Services For Flash As Cache, DAS Or Memory Caching Direct Attached Storage (DAS)

27 XtremIO RETHINKING CORE DESIGN CRITERIA Software-Defined Non-Proprietary Hardware Inherently Balanced Linear Scale-Out Architecture Flash-Optimized Sub-millisecond Data Services

IOPS Copyright 2013 EMC Corporation. All rights reserved. IOPS 28 XtremIO Scale-Out Architecture Delivers Unparalleled Functional IOPS 4K Random Read Performance 1,250,000 4 X-Bricks 1,000,000 400,000 4K Random Write Performance 4 X-Bricks 750,000 500,000 2 X-Bricks 300,000 200,000 2 X-Bricks 250,000 1 X-Brick 100,000 1 X-Brick 0 0 Sub-Millisecond Response Time With All Data Services Operating

29 XtremIO Data Services Real-time, full-speed, no limitations Real-time, inline deduplication Space-efficient snapshots Volume clones Thin provisioning Full VMware VAAI integration Flash-specific data protection

31 XtremIO For Databases Production on flash Over 130TB usable performance storage Full time administrator Massive storage footprint Can t quickly scale or respond

32 XtremIO For Databases 10X Better Performance 90% Less Floor Space 80% Less Power Usage 50% Lower Cost 20% Higher Productivity

33 VDI Breakthroughs With XtremIO Simple Administration Do Whatever You Want Dramatic Storage And Server Cost Reduction Macbook Air -like SSD User Experience Near Instant Desktop Rollouts Scale To Any Number Of Desktops & I/O Per Desktop

XtremIO VDI Breakthroughs Copyright 2013 EMC Corporation. All rights reserved. 34 Deploy Ready To Use Desktops In Seconds Minimal Flash Capacity For Each New Desktop Mix Linked And Full Clones As Needed Unparalleled Desktop User Experience Extremely Attractive $/Desktop

36 EMC CAN HELP YOU LEAD YOUR TRANSFORMATION

38 CMA Customer Profile Oracle RAC Data Warehouse Challenge Extreme Performance (6GB/S Large Block Random Workload) High Oracle Licensing Costs Storage OpEx Complex Setup/Tuning Solution XtremIO All-Flash Array Application Largest Medicaid Analytics Database (Oracle RAC) In US High User Concurrency Fast Response Time Needed BRIAN DOUGHERTY Chief Data Warehouse Architect Using XtremIO we are able to engineer our Oracle RAC systems to be faster, more scalable, and handle more concurrent users, while at the same time requiring only 20% of the previous footprint in our data center Results 75% Reduction In Storage Configuration Time Reduced Space/Power/Cooling Footprint Reduction In Server Size/Processor Count & Oracle Licenses ~1.9GB/S Bandwidth Per X-Brick At <1ms Latency

Boston Scientific Customer Profile Full Clone VDI With XtremIO Copyright 2013 EMC Corporation. All rights reserved. 39 Challenge Large-scale VDI Environment Full Clone Persistent Desktops Mandatory VDI User Experience Complaints Storage Cost, Complexity Solution XtremIO One X-Brick For Desktops Application VMware Horizon View VDI VMware vsphere Hypervisor CRAIG ENGLUND / SEAN COLLIER Principal Virtualization Architect/Sr. Administrator There are few things in history that have a significant impact on advancing technology I see XtremIO as one of those technologies. Results Incredible End User Desktop Experience Lower $/Desktop Than Prior Storage Array Rapid Desktop Rollouts During Production Hours Complex VDI Made Simple And Successful