Ejecución de sizer para SimpliVity Partners

Customer end

1. Practical Sizing of OmniStack Solutions … Starting Shortly

2. Practical Sizing of OmniStack Solutions SimpliVity UK & Ireland Technical Webinar Series – June 24 th 2016 June 2016 Stuart Gilks, SimpliVity Solutions Architect Director Antony Perry , SimpliVity Solutions Architect – Channel

3. VOIP Issues? Alternative access via telephone call-in: UK [phone] or Ireland [phone] Full list at http://bit.ly/1kOohYr Questions? Please use the Q&A facility in WebEx We will cover Qs live, on chat or if involved via follow up We record these webinars Please share with colleagues Use as a resource for those who registered but couldn’t attend Feedback Topics you want us to cover Things we can improve Things you find most valuable Use Q&A or email [email protected] Webinar Logistics

4. An Intro: UK SA Webinars SimpliVity Solution Architecture/Process recap Practical Sizing - unplugged Sizing Examples New Reference Collateral Q&A Forthcoming Sessions and Events Close Agenda

5. Congrats!!

6. Strengthen SA relationships with our UK partners Understand how to win with SimpliVity Understand tools, processes and procedures Differentiated from PartnerAdvantage Insight to available collateral Your opportunity for training and enablement input !! An Intro: UK SA Webinars

7. Data Assessment Data Gathering Architectural Design Design Review 1 2 3 4 SimpliVity Approach To Solution Architecture Introduction to the Methodology Collect data from your customer Apply guidelines and formulas 2 Build the solution for the customer Perform a review of your design

8. GMTS – G uidelines, M ethodology, T ools, S haring SimpliVity Approach To Solution Architecture https:// goo.gl/aL3RX7

9. RVtools DPack https://goo.gl/9lhBz9 SimpliVity Approach To Solution Architecture

10. Enter your collected information and variables into the Data Assessment Worksheet SimpliVity Approach To Solution Architecture 3. Compute Node Info 2. Adjust for Growth 1. CPU Memory Disk input 4. Configuration Variables

11. Review the suggested models and pricing in the heat map SimpliVity Approach To Solution Architecture

12. Practical Sizing of OmniStack Solutions Stuart Gilks, SimpliVity Solutions Architect Director

13. Agenda Data Gathering - Avoiding GIGO Better decisions working with raw data Narrowing down the options Real life examples Assuming you are at least familiar with… DPACK & RVTools Basic/traditional virtual infrastructure sizing SimpliVity architecture (SA 101) SimpliVity GMTS process

14. Quiz - Most frequently contended resource? CPU Memory Storage Networking Storage Performance Networking Other (post in chat)

15. Base Data Quality / Completeness Representative data captures Data for greenfield sites? Data for secure sites? HA & DR behaviour? New requirements (apps) Retiring or out of scope apps & datasets Planned Growth & Headroom Upgrades (apps/versions) Process changes (particularly backups)

16. Base Data Quality / Completeness (1) Representative data captures Busy periods / holidays Full participation Data for greenfield sites? Paper sizing; vendor stats; agreed approach Data for secure sites? Redaction or On-site analysis? HA & DR behaviour? HA Degraded (none/some/all resources) or n+1? DR – proportion of services to run in DR; HA in DR?

17. Base Data Quality / Completeness (2) New requirements (apps) Vendor stats / testing Retiring or out of scope apps & datasets Planned Growth & Headroom Measurement / mandatory/desirable Upgrades (apps/versions) CPU/memory footprint, data increases Architectural changes (EG more VMs, different DBs) Process changes (particularly backups) Leveraging SVT capability

18. Process Change – obvious example Existing infrastructure: Non-backup workload 800-1500 IOPS / Backup workload 5000-15000 IOPS Leveraging SimpliVity backups – workload requirement now 1000 IOPS

19. Supporting detail around the raw numbers… CPU Considerations Storage & Networking Gotchas Scoring Bonus Points

20. Quiz - On average how old is the infrastructure you are replacing? Less than 2 years 2 - 4 years old 4 – 6 years old 6+ years old

21. Accounting for generational improvements in CPUs Less than 2 years 100GHz -> 80-100GHz Broadwell 2 - 4 years old 100GHz 70-80GHz Broadwell 4 – 6 years old 100GHz 50-80GHz Broadwell 6+ years old 100GHz 10-50GHz Broadwell Processor info available in vHost in RVTools

22. Using SPEC benchmarks to adjust for CPU differences CINT2006 Rate results Integer benchmark Sustained performance >12600 submitted results 2006-2016 product coverage www.spec.org/cpu2006 3 x HP DL380G8 = 72GHz (excluding HT) Or 1131 SPECint_rate_base2006 3 x HP DL380G7 = 120GHz (excluding HT) Or 1056 SPECint_rate_base2006 Total CPU provisioned 192GHz or 2187 SPECint_rate_base2006

23. Using SPEC benchmarks to adjust for CPU differences Example OmniStack CPU option 1 system 2 x E5-2680v4 67.2GHz 62.4GHz net of 2 cores for OVC 1240 SPECint_rate_base2006 1151 SPECint_rate_base2006 192GHz requires 3 or 4 systems 2187 requires 2 or 3 systems Don’t forget to consider apps preference for clock speed Single threaded and latency sensitive apps generally prefer high clock speed lower core counts EG E5-2667v4 (8 cores @ 3.2GHz) vs E5-2699v4 (22 cores @ 2.1GHz)

24. Storage & Networking Beware in-guest presented storage (not reported by RVTools /DPACK) Dedupe & Compression rates assume optimal migration Remote site networking Data Protection & DR Requirement

25. Quiz - How much info do you typically have on data protection needs ? RPO Nothing specific Vague "up to X" Agreed tiers of SLA Retention Nothing specific V ague "up to X“ A greed tiers of retention Change rate Nothing specific E stimates based on use case Actuals based on snapshot consumption or incremental backup sizes?

26. Narrowing down options – common choices 2 node vs 3 node DCs 2 nodes: No 10 Gig networking required (unless using Compute nodes) 2 nodes: Node down loses 50% of resource (vs 33% for 3 nodes) Platform - Large vs Medium 20% more storage capacity but ~half the latency & more IOPS Not storage capacity/IOPS limited? Use Broadwell &/or Compute Nodes to optimise cost 2 metro/campus sites 10GigE <1ms RTT -> Stretched cluster option Scale in pairs 3 rd location for arbiter

27. DPACK – Aggregate & individual resource utilisation data (virtual & 2 physical reports) & RVTools on VM estate On-site requirement, 2 nd site or cloud for standby DR 75 Virtual Machines, currently running on VMWare 5.5 Aged physical servers would be updated as part of P->V (+10 VMs) Apps include SQL Server, Oracle (performance important), F&P, order mgmt. ~100GB data, sales audit ~500GB Consuming 7 of 9 racks in primary site aiming to reduce (substantially) Growth: No substantial new apps identified Growth rate/requirement not identified Data Protection Currently daily backups with DR for small % of VMs (& >24hr RTO) HA at primary site, ideally n+1 but DR site could be used for some capacity if RPO/RTO is low. DR at 2 nd site connected via 1Gbps Metro network 10 % of VMs are critical (hourly backups); next 40% every 4 hours; Remaining 50% daily backup is ok 9. Change Rate: no info but “not high” Example 1 – Design & Sizing Inputs

28. Current CPU 208 cores 506GHz available / 210GHz peak used 3091 SPECInt_rate_base2006 available / 1302 peak used CPU - Broadwell Equivalent Cores – 8 OmniStack systems with dual 14 core CPUs GHz – available 8 OmniStack systems GHz – peak 4 OmniStack systems SpecInt_rate_base2006 available 2.5 OmniStack systems SpecInt_rate_base2006 peak 1 OmniStack system Memory – 1081GB available, 623GB used 1 SVT node @ 768GB configured (668GB usable) satisfies used 2 SVT nodes @ 768GB is +20% vs current available 3 SVT nodes @ 512GB satisfies n+1 for used & is +20% vs current available Sizing Detail (1)

29. Storage 19.3TB current capacity 13.4TB primary data IOPS 20,000 peak 10,000 excluding backup 4,000 sustained Large platform preferred for latency 2 Large nodes 35% full 2 Medium nodes 41% full 3 Medium nodes 28% full IE Plenty of room for DP & growth Networking / Remote Backups Plenty of bandwidth Space for all VM backups in 2 nd site Space & Power Currently 7 racks occupied by ~24 servers, storage, networking etc Estimated power draw of existing infrastructure of 9kW (+13.5kW cooling @ PUE of 1.5) Sizing Detail (2)

30. Primary Site 1) 3 x CN3400 OmniCube / medium platform OmniStack 2) 2 x CN5400 / large platform OmniStack 3) 2 x CN3400 / medium platform OmniStack Secondary Site Single CN3400 / medium platform OmniStack (2U) Space/Power 6U/2kW or 8U/2.6kW Matches or exceeds all current resources & functionality Full HA at primary site plus provides highly efficient DR Space saving: 96-97% (7 racks -> 6 or 8U) Power Saving: 71-78% power & cooling saving (>50 tonnes of CO 2 pa) Complete solution management solely through vCenter with VM granularity Ability to significantly improve RPO/RTO with SimpliVity’s HyperGuarantee Cost savings through VMware licence re-purposing/retiring (16 sockets -> 6 or 8) Plus usual SimpliVity benefits Performance acceleration through DVP Single pane of glass management Integrated no performance penalty backups VM-centric policy driven mgmt Preliminary Design - Best n+1 solution, requires 10Gig n/w - Best latency - Most cost effective

31. 106 Servers – 85% physical, storage on NetApp FAS (NFS, iSCSI and FC) Production, isolated Test & BCP (DR @ 2 nd site) 11 ESXi hosts running VMWare 5.5 (6 Prod, 3 BCP, 2 Test) – DPACK & RVTools Apps include SQL Server (predominant), Small F&P, sharepoint & misc others Consuming 10* racks across both sites, relocating 2 nd site Lots of concern from DBAs on moving physical to virtual Growth: Size for current P+V, growth through additional purchase Data Protection Currently netapp snapshots + daily backups SQL Dump and copy to 2 nd site HA for all, n+1 for prod/DR DR at 2 nd site connected via 4x1Gbps Metro network (maybe 10Gbps in the future) 9. Change Rate: no info initially Example 2 – Design & Sizing Inputs

32. 90+ physical servers “Islands of resource” mentality with DB Admin & Architects Representative sizing data Representative samples from Physical servers Representative sample / Peak syncing across disparate physical estate Physical SQL Servers hugely over-provisioned (vs utilised) Datasets pruned to fit within processing windows – looking for performance improvements to extend datasets NetApp FAS served HPUX via FC as well as wintel platforms Route forward? Joint PoC test for virtual & P->V workloads Selected key SQL Server workloads to benchmark (current prod, current physical, SVT) Encourage extensive testing via CTS and then on-site PoC (RAS, VM density, Performance, P->V) SVT lead, open book shared design Example 2 – Complications

33. Current CPU 60 HP BL465 G1-G8 50 HP DL385 G2-G8 4 HP DL585 G2 1200 cores / 3200GHz / 17500 SPECInt_base_rate2006 Available CPU – Haswell Equivalent Cores ~46 OmniStack systems GHz ~52 OmniStack systems SpecInt_rate_base2006 16-24 OmniStack systems Reality – peak CPU utilisation very low on physical systems Available Memory – 3.8TB available, pretty much all used* About 6 nodes @ 768GB configured Storage 110TB allocated with high capacity utilisation ESXi Servers total IOPS ~2000 SQL Servers are very peaky but SQL Dumps often the spike Sizing Detail (1) – Aggregate level

34. PoC – SQL Server Performance Testing (CN3400) Key demanding workloads on key platforms selected Repeatable response time and completion time for tasks - Baseline on current infrastructure (HP/10kSAS/W2k8/SQLServer2k8) - New equiv Physical host - SimpliVity testing with minimal adjustment - SimpliVity testing with W2k12/SQLServer2k12 - SimpliVity testing with SVT+MS+VMware best practices VM Rightsizing of P->V of 16 vCPU, 128GB RAM 16vCPU / 64GB RAM 8 vCPU / 32GB RAM 4 vCPU / 16GB RAM VM Density S/M/L VM templated profile (CPU, Memory, IOPS) Scale testing to 120 Server VMs per node (4x current density) Sizing Detail (2) 240 mins 120 mins 100-110 mins 99-119 mins 52-58 mins 100 mins 104 mins 108 mins

35. CN5400 based production (lowest latency for SQL Server apps) Primary Site 4 x CN5400 - 24 core, 512GB RAM* DR/BCP 4 x CN5400 Test 3 x CN2400 – 8 core 256GB RAM Yr3 addition +4 nodes (2+2) Space/Power 90% space and 95% power savings Compute node for FC-attached tape-out to TSM (core licensed) Tape backup reduced 97% Potential stretch-cluster reconfig post-DC move (DC location dependent) Preliminary Design

36. Sizing Q&A

37. 2015 State of Hyperconverged Infrastructure EMEA 38 Partner Portal https :// goo.gl/DW6kD4

38. Placeholder for customer roundtables

39. For topics that you want covered in the Solution Architect Level 201 training , email [email protected] . Feedback for Level 201 Partner Training

40. A B C D E F G Customer Vertical Use Cases Partner Competition Date Link to Recording 1 Francis Drilling (4 x OmniStack, 2 Sites) Oil&Gas, Logistics DCC, DP/DR, UCS win Cisco UCS IBM, Nutanix 8/6/15 Francis Drilling Fluids 2 Verso Paper (19 x OmniCubes, 4 Sites) Manf. DCC, DP/DR, Core Apps (Oracle, SQL) OmniCube EMC, HP 9/8/15 Verso Paper 3 Colby College (7 x OmniCubes, 2 Sites ) Education DCC, DP/DR, Critical apps (Oracle, SQL) OmniCube Dell, Nutanix 10/15/15 Colby College 4 Marketing Innovators (4 x OmniStack, 1 Site) Business Services DCC, Data Migration, Critical apps (SQL), Test/Dev, Cisco Cisco UCS Nutanix, Nimble 11/17/15 Marketing Innovators 5 City of Arvada (14 x OmniCube, 2 Sites) City Gov’t. DCC, DP/DR, Core Apps (911 Public Safety) OmniCube Dell, Nutanix 12/10/15 City of Arvada 6 Woodman’s Food Market (20 x OmniCube, 18 Sites) Retail, Food Unified Protected ROBO OmniCube Dell 3/8/16 Woodman's Food Market 7 Land O’Lakes (4xOmniStack, 2Sites;Expanding to 4 Sites) Global Agriculture Unified Protected ROBO, Tech Refresh, DP/DR Cisco UCS Dell 4/5/16 LandOLakes 8 Merlin Entertainments (45 OmniCubes, 3 Global Regions over 6 Countries, e.g. Legoland CA, Madame Tussauds ) Global Entertainment Unified Protected ROBO, Tech Refresh, DP/DR OmniCube HP, IBM, EMC DD, SYM 4/19/16 Merlin Entertainments 9 Blundstone (4 x OmniCube, 2 Sites, ~700km apart, Australia) Retail DCC, Modernization, DP/DR OmniCube HP, SYM, Veeam 4/21/16 Blundstone 10 Romac Industries (3xOmniStack & 4xOmniCube, 2 Sites,200 miles) Manf DCC, DP/DR, Tier1 Apps (Oracle, SQL, SolidWorks…) Cisco UCS OmniCube EMC, Dell, Nutanix 5/10/16 Romac 11 Guelph Hydro Electric (4xOmniStack C240M4, 2 Stretch -cluster, 5km) Utility VDI (~60), XenApp , DCC, SQL, Billing Apps, Tier1, 50 VMs NVIDIA, AutoCAD, ERSI Cisco UCS Dell EQL, NTAP, Nutanix 6/7/16 Guelph Hydro Customer Roundtable Series Key: DCC = Data Center Consolidation DP/DR = Data Protection, Disaster Recovery

41. http://demand.simplivity.com/webinars

42. Previous UK SA Webinars – Content Available 43 https:// goo.gl/KTBBDq

43. Watch out for…..UK SA Webinars 44 Friday, 24 th June 2016 10:30  Practical Sizing of OmniStack Solutions - Click HERE to register Friday 8 th July 2016 10:30 Making PoCs Rock! What’s made our PoCs successful - Click HERE to register Friday 22 nd July 2016 10:30 Deep dive to SimpliVity automation – Give your minds a REST… - Click HERE to register Friday 5 th August 2016 10:30 Upgrade in the wild. Insight to SimpliVity 3.5 customer transitions - Click HERE to register