Joe Clabby President Clabby Analytics

1. Joe Clabby President Clabby Analytics The Advantages of Heterogeneous Workload-Optimized Information Systems Environments

2. Clabby Analytics? IT research/analysis firm 2 people; 30 years IT experience (each) in systems/storage/mgt. International focus (based in Dubai last year � regularly travel the world) Last year: Philippines, Malaysia, Kuala Lumpur, India, Russia, Switzerland, Germany, Romania, Italy, Egypt, Qatar, Dubai, Canada, & all over the U.S. Previous years: China, Brazil, Taiwan, South Africa � Case Studies; Reports; Speeches/Seminars

3. Agenda What is workload optimization � and why does it matter? Define & illustrate Which applications belong where? System design characteristics Why heterogeneous choices are important Homogeneous vs. heterogeneous Example: Oracle versus IBM Summary Observations

4. What is �Workload Optimization�?

5. What is Workload Optimization?

6. What is �Workload Optimization�? Workload optimization is about matching applications characteristics/service level requirements to the best execution engines Applications get the services they need Systems provide optimized access To meet memory requirements To meet I/O requirements To meet scalability requirements To meet security requirements To meet storage requirements �

7. What is �Workload Optimization�? Workload optimization (cont�d) �Pre-integrated� is not just about packaging Its about innovation and integration at all levels Example: IBM System x Hardware: eX5 (innovation) memory scaling (access to 600 percent more memory than other x86 vendors) Allows virtualization to perform faster � increases utilization leading to higher ROI Applications and databases run faster Allows more data in memory � enabling a new generation of business analytics, decision support, and other data intensive applications Hardware and software are integrated from top-to-bottom

8. What is �Workload Optimization�? Workload optimization (cont�d) �Pre-optimized� Systems bundled and tuned for major performance advantages Example: Financial Industry/Retail Floating point as a software function IBM POWER 7 � floating point at the chip level leading to 4X to 16X execution improvements Examples: IBM zIIP and zAAP engines Specialty processors that allow Java applications or databases to execute at record rates

9. Which Applications Belong Where? Application/System buying criteria and other considerations http://www.clabbyanalytics.com/uploads/Sun_to_Mainframe_Rev_2.pdf http://www.clabbyanalytics.com/uploads/x86_Buying_Criteria_Advisory_Article.pdf http://www.clabbyanalytics.com/uploads/SystemxVirtualizationFinal.pdf http://www.clabbyanalytics.com/uploads/BusinessResiliencyFinal__2_.pdf

10. Systems Design Characteristics And Then There Were Three: x86 Multi-cores, POWER, and z Systems market is consolidating Itanium and UltraSPARC are on the outs Big differences in each architecture Scalability characteristics Consolidation, Virtualization, Provisioning and Workload Management Etc. http://www.clabbyanalytics.com/uploads/ServerMarketViewMarch2010UPDATE.pdf

11. System Design Characteristics (What I Think Is Happening)

12. Why Heterogeneous Choices Are Important An incredible amount of data is now being generated Amount of data captured is doubling or tripling every year Mobile devices, instruments, etc. This data needs to be processed and managed New applications (business intelligence, decision support, etc.) The current model of adding people to manage systems and data cannot be sustained We�re going to have to become SMARTER in how we manage our systems environments and related data

14. How Oracle and IBM Differ Contrast a homogeneous vendor vs. a heterogeneous vendor Could have chosen HP, Dell, Fujitsu�, but Clabby Analytics thinks Oracle and IBM will be the majorcompetitors in the high-end database server business Oracle�s value prop is: We build high-performance servers optimized for our stack One system (easier implementation, management) IBM�s value proposition is: You cannot afford inefficiencies in your systems (market rate of change) Systems must be optimized for particular workloads. �If you implement a system that is not optimized for your workload, those inefficiencies are only going to compound as your data continues growing unabated Like Oracle, IBM also promises easier implementation and shorter time-to-value (thanks to integrated hardware / software systems) But more flexibility to offer solutions for a client�s particular situation (multiple stacks)

16. Summary Observations Application characteristics should dictate systems choices Platforms are optimized differently Different systems offer different service levels Learn how systems are balanced Be aware that the applications/server market is changing x86 multi-cores are a game changer Sun and Itanium are fading � but POWER and x86 multi-core (and successors) are trying to move into the mainframe space Know the difference between a general purpose server and an application server

17. Summary Observations (cont'd) Understand other important heterogeneous systems differences Pre-integration is not just packaging HW and SW integration and innovation at all levels Pre-optimization Systems bundled and tuned for MAJOR performance improvements For more information, look at some of the reference materials that I have provided All are available for free at www.clabbyanalytics.com