Trends in SAP Data Centers

658 views

Published on

Realtech Consulting Webinar about SAP Migrations worldwide
- Migration source platforms are overwhelmingly UNIX (more than 90%)
- Migration target platforms are Linux (58%) and Windows (27%)
- No Limits in terms of SAP workloads to be migrated

Published in: Technology
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
658
On SlideShare
0
From Embeds
0
Number of Embeds
8
Actions
Shares
0
Downloads
52
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Trends in SAP Data Centers

  1. 1. Driving Value With ITThe Trend from UNIX to Linux in SAP DataCenters: Large. Critical. Beyond Limits.
  2. 2. U2L 3rd – General Results - Contents What Happened Since  Statistics Reloaded  Development of Price / Performance Ratio  Virtualization Reviewed  Green IT Reviewed Large. Critical.  Large Systems  Large Customers & Large Migration Projects  High Availability & Other Forms of High Criticality  Other Forms and New Understandings of HA and Criticality Beyond Limits.  Beating Lower Limits  Pushing Upper Limits Final Conclusions
  3. 3. U2L 3rd – General Results - Contents What Happened Since  Statistics Reloaded  Development of Price / Performance Ratio  Virtualization Reviewed  Green IT Reviewed Large. Critical.  Large Systems  Large Customers & Large Migration Projects  High Availability & Other Forms of High Criticality  Other Forms and New Understandings of HA and Criticality Beyond Limits.  Beating Lower Limits  Pushing Upper Limits Final Conclusions
  4. 4. U2L 3rd – What Happened Since: Statistics Reloaded (1)  UNIX is the by far most frequent migration source with more than 90%.  Other sources almost negligible and usually part of merger & acquisition or “political projects”
  5. 5. U2L 3rd – What Happened Since: Statistics Reloaded (2)  Targets comprise all architectures except mainframe-style but with widely different shares and for strongly different reasons.  x86_64 (x64) is the big winner in the field.  Equal target shares of 7% for AIX on Power and Solaris on SPARC have strongly different implications.
  6. 6. U2L 3rd – What Happened Since: Statistics Reloaded (3)  UNIX loss has gotten worse to overall 77% with HP-UX largest single loser at 46% loss rate.  Only x64-based operating systems win market share on a global scale with Linux at 56% and Windows at 23%.
  7. 7. U2L 3rd – What Happened Since: Statistics Reloaded (4)  Linux gains best in Europe (80% targets) with the largest infrastructures and projects.  x64 has conquered the Americas as exclusive migration target with Linux at 37% and Windows at 63% - overall source customer base has fewer SAP systems and smaller UNIX infrastructures: Migration to Windows frequently leads to homogeneous landscape.  Linux is weak so far in Asia-Pacific region (only 9% targets) – market is dominated by Windows (69%) – even UNIX is stronger there than Linux.
  8. 8. U2L 3rd – What Happened Since: Statistics Reloaded (5)  95% of Linux targets are UNIX sources, again with HP-UX largest single loser at 58%.  5% Windows as a result of mostly “political projects”  Mainframe architectures have disappeared as Linux targets (usually Windows instead).
  9. 9. U2L 3rd – What Happened Since: Statistics Reloaded (6)  There is movement in the database sector as well:  Oracle loses over 20% market share with accelerating tendency.  In the all targets evaluation and globally, the big winner is Microsoft SQL-Server (+ 19%).  On Linux targets, DB2 LUW gains 13% market share (overall +3%)  MaxDB has become very static (no wins, almost no losses)
  10. 10. U2L 3rd – General Results - Contents What Happened Since  Statistics Reloaded  Development of Price / Performance Ratio  Virtualization Reviewed  Green IT Reviewed Large. Critical.  Large Systems  Large Customers & Large Migration Projects  High Availability & Other Forms of High Criticality  Other Forms and New Understandings of HA and Criticality Beyond Limits.  Beating Lower Limits  Pushing Upper Limits Final Conclusions
  11. 11. U2L 3rd – What Happened Since:Development of Price / Performance Ratio (1) Basically, same rules applied as in first two evaluations New reference benchmark (ECC 6.0 EHP 4 UC)  Don’t compare US-$/SAPS with older evaluations! Basic results of first two evaluations confirmed, some tendencies clearly stronger:  Top 12 in price performance are all x64 CPUs and as low as 0.29 US-$/SAPS (AMD) or 0.36 US-$/SAPS (Intel)  Power only UNIX CPU acting in competitive range (best result 0.59 US-$/SAPS for PowerBlade)  Itanium and SPARC far off the mark and not competitive at all (with 2.48 US-$/SAPS respectively 7.72 US-$/SAPS as best results)  HP and Oracle avoid being comparable by not providing up-to-date 2-tier-SD-benchmarks for Itanium and SPARC.  Important factor for absolute cost of your data center: How much will you pay for the SAP computing power you need?
  12. 12. U2L 3rd – What Happened Since:Development of Price / Performance Ratio (2) Evaluation of performance per socket:  Power has best performance per socket, followed by Intel.  This criterion is clearly dominated by Power and x86 Intel (top 18 only from those two CPU families).  For unknown reasons, Dell falls off in benchmarks with identical x64 CPUs.  SPARC and Itanium far off, as usual Important factor for server design: Will you need 2-socket-servers, or 4-socket-servers, or more? Evaluation of performance per core:  Power has clearly best performance per core (top 3), followed by Intel (subsequent top 17), thus this criterion is again clearly dominated by Power and x86 Intel.  AMD CPUs very weak in this criterion, SPARC even worse Important factor for licensing cost in core-based database licensing models  Don’t purchase these architectures if you pay DB licenses per core!
  13. 13. U2L 3rd – What Happened Since:Development of Price / Performance Ratio (3) Evaluation of performance per thread:  Intel has clearly best performance per thread (top 7), followed by AMD (1).  Power relatively weak compared to previous criteria  Again, forget about Itanium and SPARC. Important criterion for batch processing performance and JAVA  possible explanation for some Power7 customers experiencing problems on batch performance after Power7 switch Absolute server performance:  Only criterion with SPARC on #1 position, but … … it takes 64 SPARC CPUs to beat 8 Intels and … … all x64 2-socket servers easily beat the 4-socket SPARC one. Important: x64 servers with two and four sockets deliver sufficient performance for 99% of all conceivable single SAP applications.
  14. 14. U2L 3rd – General Results - Contents What Happened Since  Statistics Reloaded  Development of Price / Performance Ratio  Virtualization Reviewed  Green IT Reviewed Large. Critical.  Large Systems  Large Customers & Large Migration Projects  High Availability & Other Forms of High Criticality  Other Forms and New Understandings of HA and Criticality Beyond Limits.  Beating Lower Limits  Pushing Upper Limits Final Conclusions
  15. 15. U2L 3rd – What Happened Since: Virtualization Reviewed Virtualization has become a commodity:  Everybody does it!  There are various solutions for x64 in general and specifically for Linux, namely VMware and KVM.  CPU and I/O throughput losses have come down to well acceptable levels.  Current technology allows for critical and high-performance SAP systems to run in virtualized environments. The progress on virtualization for x64 is another problem for UNIX – it decouples the number of servers from the number of OS partitions you need and lets you tailor systems to actual requirements.
  16. 16. U2L 3rd – General Results - Contents What Happened Since  Statistics Reloaded  Development of Price / Performance Ratio  Virtualization Reviewed  Green IT Reviewed Large. Critical.  Large Systems  Large Customers & Large Migration Projects  High Availability & Other Forms of High Criticality  Other Forms and New Understandings of HA and Criticality Beyond Limits.  Beating Lower Limits  Pushing Upper Limits Final Conclusions
  17. 17. U2L 3rd – What Happened Since: Green IT Reviewed  General rule: Better environmental specs result from development progress  CPU architectures with little or no development will lose (see chart above).  Power and x64 from Intel or AMD each have individual strengths regarding energy efficiency, e.g. AMD CPUs have lowest average waste heat factor, Power has highest absolute performance, and Intel processors average those qualities best.  This aspect is just another nail in the coffins of Itanium and SPARC.
  18. 18. U2L 3rd – General Results - Contents What Happened Since  Statistics Reloaded  Development of Price / Performance Ratio  Virtualization Reviewed  Green IT Reviewed Large. Critical.  Large Systems  Large Customers & Large Migration Projects  High Availability & Other Forms of High Criticality  Other Forms and New Understandings of HA and Criticality Beyond Limits.  Beating Lower Limits  Pushing Upper Limits Final Conclusions
  19. 19. U2L 3rd – Large. Critical. – Large Systems Single system limits said to be critical boundaries have generally been beaten in the x64 world: REALTECH has implemented, migrated and/or evaluated systems running Linux on x64 CPUs with …  … more than 120,000 IOPS,  … more than 200,000 SAPS,  … 512 GB of memory,  … databases up to 20 TB in size.  More important: All these values are not the end. More is absolutely feasible. Limits we found in certain scenarios were actually caused by the design of the SAP application stack. Other OS did the same – Linux is nothing special any more, in a positive sense. But this also means: Linux doesn’t get special considerations concerning support any more. It’s supposed to work and to be well supported! However: Both SAP and some vendors still struggle with professional enterprise-style support for Linux and / or their products on and for Linux.  Make sure you have good contracts and good contacts!
  20. 20. U2L 3rd – General Results - Contents What Happened Since  Statistics Reloaded  Development of Price / Performance Ratio  Virtualization Reviewed  Green IT Reviewed Large. Critical.  Large Systems  Large Customers & Large Migration Projects  High Availability & Other Forms of High Criticality  Other Forms and New Understandings of HA and Criticality Beyond Limits.  Beating Lower Limits  Pushing Upper Limits Final Conclusions
  21. 21. U2L 3rd – Large. Critical.Large Customers & Large Migration Projects (1) Generally: The largest customers in almost all industries have approached the x64 and Linux topic for their SAP systems, e.g.  the world’s largest reinsurance  the world’s largest chemical company  some of the world’s largest airlines  some of the world’s largest banks  some of the world’s largest car manufacturers  some of the world’s largest telecommunication companies  some of the world’s largest pharmaceutical companies They all started as large UNIX shops. Those big ones who haven’t tackled the topic yet will not be early adaptors with an advance to competition, but mainstream at best. Some even may be late!
  22. 22. U2L 3rd – Large. Critical.Large Customers & Large Migration Projects (2) Large U2L migration projects are feasible and controllable … if you don’t fall for the following typical organizational mistakes:  Lack of planning, or unrealistically tight project budget, or both  Lack of time to accomplish all necessary tasks  Total lack of or significant shortage in internal and external project management  Missing capacity of important internal or external project staff members during the project, including too few certified migration specialists  Missing acceptance of the new OS or DB with internal staff members, along with limited knowledge of the new environment  Solutions do not fit management expectations and have not been evaluated thoroughly before the final “Go”  Missing technology components are “worked around” instead of buying necessary licenses  Avoid workarounds – at all cost!
  23. 23. U2L 3rd – Large. Critical.Large Customers & Large Migration Projects (3) Large U2L migration projects are feasible and controllable … if you don’t fall for the following typical technical mistakes:  Underestimating the complexity of the target environment  Not using the “best of breed” approach – especially the “one to one mapping” mistake  Implementing functionality in wrong layers of the architecture, instead please …  … leave infrastructure functionality in infrastructure layer(s)  … leave application functionality in application layer(s)  Giving key components too high ratings regarding robustness and functionalityLinux and UNIX are very similar to each other, but don’t mistake them to be identical. They are not!
  24. 24. U2L 3rd – Large. Critical.Large Customers & Large Migration Projects (4) Successful migration projects feature the following organizational & technical specs:  Realistic and careful planning  A feasible timeline  Experienced project management on both sides  A knowledgeable migration team  Sufficient budget and manpower  Comprehensive testing of all technical components in the customer’s infrastructure environment prior to important milestones. And even then, problems may, no, will occur. Be prepared for them. The good news is: If your general approach was right, you will be able to solve them!
  25. 25. U2L 3rd – General Results - Contents What Happened Since  Statistics Reloaded  Development of Price / Performance Ratio  Virtualization Reviewed  Green IT Reviewed Large. Critical.  Large Systems  Large Customers & Large Migration Projects  High Availability & Other Forms of High Criticality  Other Forms and New Understandings of HA and Criticality Beyond Limits.  Beating Lower Limits  Pushing Upper Limits Final Conclusions
  26. 26. U2L 3rd – Large. Critical.HA & Other Forms of High Criticality (1) First things first: “High Availability is a way of life, not a collection of tools.”  You need to implement HA as a process your IT works according to. There are tools who may help you to do so, for SAP on Linux specifically the following:  SLES High Availability Extension (includes Pacemaker, Corosync, OpenAIS)  High Availability for RHEL (currently still rgmanager-based, thus technically not up-to-date)  SteelEye Lifekeeper, a viable 3rd party solution  Veritas Cluster, very comprehensive 3rd party solution recommended by Red Hat for “big picture approaches”  Some other cluster tools REALTECH does not want to actively promote for various reasons. All have specific and individual advantages and drawbacks, and associated costs. Due to our experience in past projects, REALTECH will implement cluster-based HA on Linux only after a preceding PoC specific for a customer’s infrastructure.
  27. 27. U2L 3rd – Large. Critical.HA & Other Forms of High Criticality (2) First things first: “High Availability is a way of life, not a collection of tools.”  You need to implement HA as a process your IT works according to. Other forms of HA & High Criticality handling comprise  Database-managed clustering, e.g. Oracle RAC or DB2 Pure Scale  Application-layered availability handling, specifically SAP Enque Replication  Usage of virtualization-layer mechanisms for continuous availability  High Availability by multiple redundancy Our experience and advice to customers: Find & build the optimum mix of the various options for your environment.
  28. 28. U2L 3rd – General Results - Contents What Happened Since  Statistics Reloaded  Development of Price / Performance Ratio  Virtualization Reviewed  Green IT Reviewed Large. Critical.  Large Systems  Large Customers & Large Migration Projects  High Availability & Other Forms of High Criticality  Other Forms and New Understandings of HA and Criticality Beyond Limits.  Beating Lower Limits  Pushing Upper Limits Final Conclusions
  29. 29. U2L 3rd – Beyond Limits. – Beating Lower Limits (1) Initial situation and motivation: Review of the platform strategy for SAP of a large health sector enterprise  Achieve future-proof alignment for at least the next five years.  Evaluate and prove the feasibility of a full (migration of databases) or partial (Linux application servers) conversion from Power to x86.  Review and financially estimate the technical, organizational and economical impact of the suggested steps.  Cost optimization: Increase overall IT capabilities at equal or lower costs.  Include parallel rebuild of data centers.  Compare IT with competitors in the sense of an audited benchmark.  Create preconditions for a quick introduction of a “Private Cloud“ – without having to implement the step right away
  30. 30. U2L 3rd – Beyond Limits. – Beating Lower Limits (2) Basic assumptions for the design recommendations  Identical but separate landscapes for test and production  Standardization  Customer standards are compulsory (here Oracle as database, VMware for virtualization)  Use commodity HW if possible to leverage administrative effort and achieve a high degree of flexibility (reallocation)  Fast & simple provisioning processes  Maintain as much existing know-how and operational procedures as possible  Future-proofness of all suggested options  High vertical and horizontal scalability  Quick and enduring availability of the hardware from several vendors if possible and of skilled experts  Consider everything as part of a high-availability environment.
  31. 31. U2L 3rd – Beyond Limits. – Beating Lower Limits (3) Possible design options  1:1 transfer of existing CI/DB landscape on  Newest Power7 (large) servers  x86 servers with SUSE Linux  Transfer SAP applications into a layered model:  Application services layer: Scalability and high-availability by multiple dialog instances  Central services layer SPOFs eliminated by Enqueue Replication and Message Server Clustering  Database services layer with Oracle RAC Scalability beyond the limits of a single bare-metal server with Oracle RAC, thus enabling rolling maintenance  This architectural approach enables changes in one layer without affecting the other, and it optimizes the core-based database licensing the customer has contracted to.
  32. 32. U2L 3rd – Beyond Limits. – Beating Lower Limits (4) Cost calculation for 1:1 transfer on Power7:
  33. 33. U2L 3rd – Beyond Limits. – Beating Lower Limits (5) Cost calculation for layered model with Power7:
  34. 34. U2L 3rd – Beyond Limits. – Beating Lower Limits (6) Cost calculation for layered model with x86/Linux:
  35. 35. U2L 3rd – Beyond Limits. – Beating Lower Limits (7) Decision and next steps  Customer goes for layered architecture with x86 blades although over 5 years there is no significant cost-advantage compared to layered architecture on Power7.  Important decision factors:  Significant vendor-lock-in with IBM Power blade solution  Homogenization of SAP data center operations due to appliances already coming with x86 & SUSE Linux  Potentially high subsequent costs for memory intense applications on Power  Standardized server procurement for Windows and Linux Current project status: Target production environment is being implemented. Planning: Migration of the complete system landscape by end of 2013
  36. 36. U2L 3rd – General Results - Contents What Happened Since  Statistics Reloaded  Development of Price / Performance Ratio  Virtualization Reviewed  Green IT Reviewed Large. Critical.  Large Systems  Large Customers & Large Migration Projects  High Availability & Other Forms of High Criticality  Other Forms and New Understandings of HA and Criticality Beyond Limits.  Beating Lower Limits  Pushing Upper Limits Final Conclusions
  37. 37. U2L 3rd – Beyond Limits. – Pushing Upper Limits (1) In another U2L migration project we managed to …  … double available SAP computing power (excluding BWAs) and …. SAPS BWA SAPS AppServer 961.200 SAPS Oracle / ZI Linux SAPS BWA / ZI HPUX 299.000 249.270 995.100 500.370 38.312
  38. 38. U2L 3rd – Beyond Limits. – Pushing Upper Limits (2) In another U2L migration project we managed to …  … simultaneously cut SAP-related server costs by over 80%.
  39. 39. U2L 3rd – General Results - Contents What Happened Since  Statistics Reloaded  Development of Price / Performance Ratio  Virtualization Reviewed  Green IT Reviewed Large. Critical.  Large Systems  Large Customers & Large Migration Projects  High Availability & Other Forms of High Criticality  Other Forms and New Understandings of HA and Criticality Beyond Limits.  Beating Lower Limits  Pushing Upper Limits Final Conclusions
  40. 40. U2L 3rd – Final Conclusions x64 and Linux have arrived in SAP data centers  Both have become mainstream solutions.  The two frequently offer the best option in price/performance ratio for large UNIX environments.  Size limits don’t seem applicable any more.  Both x64 and Linux are well suited to deliver high performance and high availability …  … but these don’t come without endeavor! Some typical mistakes in transition projects are connected to the similarity of UNIX and Linux. Some UNIXs are doomed, and the transition of large environments takes time:
  41. 41. About REALTECH SAP experts since its foundation in 1994 Headquarters: Walldorf, Germany More than 330 employees* Worldwide presence More than €39 million in revenues* Founding member of the SAP LinuxLabU2L whitepaper download: www.realtech.com/linuxInterested in our services? - Contact us: customer-services@realtech.comWant to talk to SUSE? - Contact us: dop@suse.com* as of December 31, 2011
  42. 42. Driving Value With IT SAP Solution Manager SAP Mobile Technologie Cloud SAP HANA

×