IDC Market Update Agenda IDC Overall GDP And

IDC Market Update

Agenda Ø IDC Overall GDP And IT Economic Update Ø IDC Worldwide Server Forecasts Ø IDC HPC Market Update § Current Market Dynamics § HPC Forecast § IDC’s View on High Growth and Low Growth Segments § Summary

IDC Overall GDP And IT Economic Update

The Downturn’s Impact On IT John Gantz IDC Chief Research Officer

An Economic Shock: Worldwide GDP Forecast Source: IMF, Consensus Economics, Inc. , EIU 5

An Economic Unknown: Government Actions $1 -3 Trillion For ØBank Bailouts ØTax Cuts ØInfrastructure ØSMB Incentives ØHealth Care ØSmart Grid ØAuto Manufacturers ØGreen Building 6

Impact on IT 7

New Downside Scenario: It May Go Lower In 2009 Worldwide IT Spending PCs Servers Storage Printers Networking Software Services Source: IDC Q 3 2008 Worldwide Black Books; IDC Scenario Estimates

The New IT Budgets: Recovery in 2010 Global IT Spending Source: IDC Q 4 2008 Worldwide Black Book 9

Short Term Hot and Cold Spots Hot Cold Watch Search & Discovery Commodity Hardware Location Based Services Security Management Project Based Consulting Real Time Analytics Mobile Data Mobile Voice Virtualization Mgt SW Business Analytics IT Training Ethical Hacking IT Outsourcing & BPO Finance & Accounting IP Surveillance Saa. S ERP Smart Grid Internet Advertising Sales Force Automation Video Search Virtual Machine SW Manufacturing SW Chipless RFID Metro WDM/Wi-max DSL Reputation Mgt SW Storage Replication Large Enterprise IT MPOG Virtual Artifacts 10

Essential Guidance § The leading edge technology users will use the downturn to pull further ahead ØUsing IT as a competitive weapon § Opportunity will come from feeding/riding the technology transitions § Government actions will create new opportunities ØBut will change frequently and will vary by country 11

IDC Worldwide Server Forecasts

Computing Infrastructure Trends Matt Eastwood Group Vice President Enterprise Platforms Research April 2009

WW Server Forecast: This is a Market Reset YOY Revenue Growth 4 Q Rolling Spending ($M) -$15 B -25% Source: IDC Quarterly Server Forecast. March 2009

Worldwide Server Market, 1996 -2010 Revenue ($M) ‘ 97 -’ 08 Change 45% 16% 14% -$10. 7 B 31% -$8. 6 B 32% -$3. 0 B 55% 23% +$16 B 105% of market growth was in x 86 spending 15

Server Capability (and Density) Soars 16

Market Segments Continue to Shift % Server Spend Mega Datacenters Datacenter § Mega – Homogeneous – Dense/Low Cost § Small Site (SMB/Branch) – Low Cost – Ease of deployment – Ease of management 9% 14% 77% 16% 19% 65% Enterprise Datacenter § Enterprise – Heterogeneous – Legacy / Complex – Consolidating Quickly 17

The Top 10 HPC Predictions for 2009

IDC HPC Market Update

Top Trends in HPC The global economy is impacting all HPC segments HPC declined -3% for 2008 overall § 2009 is projected to decline -5. 4% – A major change from the 19% yearly growth over the previous 4 years § We are forecasting 3% growth for the next 5 years – Growth starting in 2010 Major challenges for datacenters: § Power, cooling, real estate, system management § Storage and data management continue to grow in importance Software hurdles will rise to the top for most users § Driven heavily by multi-core processors and hybrid systems

2008 HPC Server Market Results

HPC Server Market Size By Competitive Segments (2008 Data) HPC Servers $9. 77 B Workgroup (under $100 K) $2. 0 B Supercomputers (Over $500 K) $2. 7 B Divisional ($250 K - $500 K) $1. 4 B Departmental ($250 K - $100 K) $3. 7 B

Vendor HPC Market Shares In 2008: All HPC Segments Source IDC, 2009

Total HPC Revenue by Processor Type Source IDC, 2008

Total HPC Revenue by OS Source IDC, 2008

Growth In HPC Clusters

Why is Commodity Hot? . . Price! Ave. CPUs / ASP ($K) System $K / CPUs / $M x 86 50. 7 21 2. 4 419 RISC 98. 3 10 9. 5 105 EPIC 55. 9 7 7. 6 131 873. 4 12 73. 6 14 Vector Source IDC, 2009

Cluster Vendor Market Shares, 2008 Source IDC, 2009

IDC HPC Market Forecasts

Major Forecast Assumptions Ø Macroeconomic turmoil will slow the global economy, reducing overall IT and HPC server spending Ø HPC server sales will show a significant decline extending through mid-to-late 2009, but not as severe as the overall server market § High-end of the market will be more resilient to the general economic condition than other segments § Mid-to-low end of the market will still lag the enterprise profile, but will follow it more closely than before Ø Clusters will continue its penetration into HPC

Major Forecast Assumptions Ø We expect to see different impacts in different industries § Automotive and Financial sectors will hurt the most in the near term, expect rebound in 2010 § Energy sector will see stronger spending in search for alternative resources § Bio-life science will show flat to moderate growth § Gaming/movie sector will spend at healthy rate § Government, Defense and Universities will present flat to moderate growth Ø Green IT will slowly have more impact on procurement decisions Ø Petascale initiatives around the world will drive up sales in high-end of the market

New HPC Forecast on Revenue, Units, ASP, 2007 – 2013

Forecast Comparison on HPC Revenue, Pre & Post Crisis, 2007 – 2013

New HPC Revenue ($M) Base Case Forecasts, 2007 – 2013, By Competitive Segment

Down Side Forecasts ($M), 2007 – 2013, By Competitive Segment

NEW HPC Application/Industry Forecasts, 2007 - 2013 Application Segment 2007 2013 CAGR(07 -13) Bio-Sciences $1, 530, 197 $1, 781, 031 2. 6% CAE $1, 225, 638 $1, 562, 311 4. 1% Chemical Engineering $256, 033 $260, 900 0. 3% DCC & Distribution $575, 608 $835, 046 6. 4% Economics/Financial $305, 032 $421, 115 5. 5% EDA $741, 054 $948, 920 4. 2% Geosciences and Geo-engineering $587, 074 $807, 039 5. 4% Mechanical Design and Drafting $125, 891 $98, 205 -4. 1% Defense $919, 574 $1, 186, 212 4. 3% Government Lab $1, 440, 837 $1, 863, 896 4. 4% University/Academic $1, 867, 560 $2, 337, 419 3. 8% Weather $393, 261 $545, 329 5. 6% Other $108, 663 $140, 644 4. 4% $10, 076, 423 $12, 788, 066 4. 1% Total Revenue Source: IDC, 2009

IDC View On High Growth vs. Low Growth Segments For 2009

Color Coding Scheme

Supercomputers By Application/Industry Segments

Divisional By Application/Industry Segments

Departmental By Application/Industry Segments

Workgroup By Application/Industry Segments

Total Market By Application/Industry Segments

IDC HPC Summary

Major Customer Pain Points #1 Dealing With The New Economic Realities Clusters are still hard to use and manage System management & growing cluster complexity Power, cooling and floor space are major issues Third party software costs Weak interconnect performance at all levels Applications & programming — Hard to scale beyond a node RAS is a growing issue Storage and data management are becoming new bottle necks § Lack of support for heterogeneous environment and accelerators § § § §

Major Customer Pain Points Software has become the #1 roadblock § Better management software is needed – HPC clusters are hard to setup and operate – New buyers – require “ease-of-everything” § Parallel software is lacking for most users – Many applications will need a major redesign – Multi-core will cause many issues to “hit-the-wall”

Questions? Please email: hpc@idc. com Or check out: www. hpcuserforum. com

Questions? Please email: hpc@idc. com Or check out: www. hpcuserforum. com

Backup Slides

The Top 10 HPC Predictions for 2009

1. The HPC Market Will Dip In 2008 and 2009 But Will Remain a Bright Spot in the IT Space • IDC estimates full-year 2008 HPC server revenue will be about $9. 6 billion, down 4. 2% from 2007. Worldwide High-Performance Technical Systems Total Market Revenue Forecast, 2007 - 2012, Base Case • Our base case forecast shows revenue declining 5. 4% in 2009, resuming modest growth in 2010, rebuilding to robust 9. 6% growth and $11. 7 billion revenue in 2012. • HPC looks to remain one of the bright spots in the IT space. WHY? Source: IDC 2009 § HPC is mission-critical: § IDC late-2008 survey: None of the 110 government, industry, academic organizations planned to reduce HPC use in 2009. § But sites will be more conservative about new spending.

1. Fiscal Conservatism Will Affect HPC Market Segments Unequally • Some automotive and financial services firms will shrink capex even in mission-critical areas, including HPC. • In leading oil and gas companies, and in some entertainment and consumer product firms, budget cuts will be rare and HPC growth plans will usually be pursued, sometimes with delays. • Government and academia will likely follow historical patterns by reacting less quickly and deeply to the economic downturn than the private sector does. • In the U. S. and elsewhere, HPC will compete for funding with other pressing priorities.

1. The Global Recession Has Other Implications • The moderate revenue decline will trigger some vendor consolidation. • Increased focus on cost-effectiveness will make standards-based clusters even more appealing. Products that boost the efficiency of existing HPC resources will also do well. • More sites will apply simulation and analysis to existing data center designs to grow performance with minimal impact on power, cooling, and facility space. • Capex-free HPC cycles delivered via service-oriented grids (and perhaps even some via cloud computing) will become more appealing to new users and for periodic, overflow work.

2. HPC Storage Will Increasingly Outpace the HPC Server Market • The storage market will stay closer on course in 2009 as the server market dips, enabling the storage growth rate to outpace servers by at least 5%. • Some large commercial storage firms will still fail to understand the learning curve needed to sell to HPC buyers. • More hardware OEMs will start offering HPC storage solutions. Not all of them will have as much value to add as they believe. • Flash has major advantages but will be applied judiciously until costs drop.

3. The Petaflop Club Will Gain More Members • PF initiatives are under way around the globe. • There will eventually be two membership levels: § Systems with substantial custom engineering price tags at times exceeding $100 million § Megaversions of standards-based clusters sold at a fraction of that amount • 2011 -12: The petascale bake-off between the creations of the DARPA HPCS program and Japan's Keisoku Project § Will Japan blow the U. S. out of the water with superior real-world performance again, as the Earth Simulator did in 2002? § Will these innovation-packed supercomputers lead the rest of the HPC market forward to unprecedented programmability, productivity, and performance, or will they branch off as a separate species with little relevance for the mainstream HPC market?

4. HPC Supply Chain Use Will Become a Metric For Industrial Competitiveness • IDC/Co. C research: 97% of the firms that had adopted HPC said they could no longer compete or survive without it. § Another 2008 IDC/Co. C study showed that except for oil and gas, HPC isn't used much yet in supply chains. • Today's affordable HPC entry-level systems, along with SMB-oriented utility computing offerings will capture more tier 2 and tier 3 industrial suppliers beginning in 2009. • HPC use will start to become a metric for supply chain competitiveness, as it already is for tier 1 firms.

5. Power and Cooling Will See Lots of Innovation But No Major Breakthroughs • Power & cooling costs will remain a top issue for HPC sites. • “Green" savings will be earmarked for the purchase of additional performance – the holy grail of HPC. • There will be innovations in 2009 and more simulation of data centers for energy efficiency, but no major breakthroughs. • The primary "Green" driver in HPC will be to obtain more performance within a fixed or more slowly growing green footprint. In most other IT sectors the primary driver will be to reduce costs to improve profitability.

6. Competition Will Heat Up In the Alternative Processor Wars • Nearly 10% of the sites IDC interviewed for our enduser, demand-side research were using alternative processors. § Alternative processors have not yet made big market share gains in HPC, but neither are they insignificant. • Alternative processor vendors are "messaging" in their marketing campaigns that the future of HPC belongs to them. • For 2009 and the near term, x 86 processors will remain in the driver's seat. Intel's Nehalem and AMD's Shanghai processors will deliver enough advances to keep them there.

6. Competition Will Heat Up In the Alternative Processor Wars (continued) • Innovations such as Nvidia's CUDA have substantially eased the burden of programming alternative processors, but for now this burden is still more onerous than coding within the mainstream x 86 ecosystem. • In 2009, users and vendors will increase their exploration of processor and system-level heterogeneity as the processor wars heat up. • Will the majority of future HPC systems use a mix of processor types? • Will x 86 and alternative processors incorporate more and more of each other's capabilities and grow more and more alike?

7. Standard Products Will Grow, But More Codes Will See Retrograde Performance • A late-2008 IDC survey: one in eight HPC sites (12%) had some codes that ran more slowly on their newest HPC system than on the prior one. § 50% of the sites said they expected to see retrograde performance on some codes within 12 months. • The culprits: escalating core counts that exceed the codes' scalability, driven by the inability to move data in and out of each core fast enough to keep the cores busy. • Most HPC ISV applications were designed to run on one core with relatively strong access to main memory. § Multicore/manycore processors have dramatically changed the bytes-to-flops ratio. § And energy-saving, tuned-down processor speeds reduce single-threaded performance.

8. The Highly Parallel Programming Challenge Will Increase • Hardware parallelism from burgeoning core counts and system sizes is racing ahead of programming paradigms and the time available to programmers. § Manycore processors and heterogeneity add to the programming challenge. • The parallel performance "wall" will reshape the nature of HPC code design and system usage. • New DARPA HPCS languages could transform highly parallel programming starting a few years from now, but some users will resist this revolutionary change.

8. The Highly Parallel Programming Challenge Will Increase (continued) • Solutions in 2009 will be less revolutionary. They include: § Hide parallel hardware behind optimized application-specific libraries (Interactive Supercomputing, Acceleware) § Optimize intra-node performance for multicore (Acumen) § Abstract from distributed parallel hardware (Scale. MP, PGAS languages) § Address processor heterogeneity by extending the x 86 ISA within the compiler to include parallel accelerators (e. g. , Convey Computer) • Eventually something will need to give. HPC programmers will need more efficient programming paradigms, more innovative approaches for redesigning applications, more balanced hardware architectures, or all of the above.

9. Software Licensing Costs Will Become a More Universal Chokepoint • For most medium to large HPC industrial sites, application software fees already exceed the HPC server costs, often by a factor of two. § For smaller HPC users the application fees can exceed the server hardware costs by four to five times. • Software vendors have been hard pressed to develop pricing models that turn an adequate profit while enabling their customers to exploit the many additional CPU hours that rampant hardware parallelism has made available. § In 2009, software vendors will make further progress in addressing this challenge but licensing costs will become a major issue for nearly all HPC sites that rely on commercial software.

10. Ease-of-Everything Will Gain Ground At the Low End and Beyond • In 2007 and 2008, HPC vendors introduced server products priced under $50, 000 and were designed to meet that segment's "ease-of-everything” requirements. • In 2009, the underlying complexity of HPC server systems will continue to grow and ease-of-everything solutions will become more widespread in the workgroup segment. • Hardware complexity will also grow for users outside of the workgroup segment, and this will make ease-ofeverything solutions more appealing at higher price points as well.

Essential Guidance: General • 2009 will be a year of evolutionary rather than revolutionary change. • Existing major challenges will remain inadequately addressed: § § § Highly parallel programming System imbalance (the "memory wall") Power and space usage Software licensing costs Ease-of-use – dealing with the growing system complexity • The global economic recession will not significantly slow progress on these challenges or on scientific and engineering work, but it will slow sales enough to cause a dip in 2009.

Essential Guidance: HPC User/Buyers • In the difficult economy, vendors will compete more fiercely for your business. This will present opportunities to drive harder bargains that may no longer be available as the HPC market begins to recover in late 2009 and onward • Users whose capex is under greater pressure in 2009 than their opex should consider sending overflow work to utility computing providers § Some of these providers offer ISV applications and expertise as well as cycles § External cycles work best on less time-critical and less-security sensitive jobs

Essential Guidance: HPC Vendors • The economic downturn will make buyers even more price-sensitive and this will be more favorable for standards-based clusters. • Clusters will continue to put pressure on profits and challenge vendors to find new ways to deliver added value. • Industrial HPC buyers will likely complete planned R&D-driven HPC purchases over the next few months, then reduce capital spending in many industries (excluding oil/gas). • IDC expects government and university HPC purchasing to enter a flat growth period, with purchasing delays possible for the first six months or so of the new U. S. Administration.

Essential Guidance: HPC Vendors • National security and homeland defense operations will continue to develop additional requirements for HPC systems • New applications areas for HPC may be developed based on database and pattern matching requirements • IDC expects R&D for alternative energy sources, as well as nuclear, coal, climate modeling, and oil/gas to be strong growth segments

Conclusions • 2009 will be a year of evolutionary rather than revolutionary change in the worldwide HPC market § Incremental advances will alleviate but not resolve persistent issues, such as highly parallel programming, power and cooling costs, and software licensing costs • IDC predicts the HPC market will fall another 5. 4% in 2009 before resuming modest growth in 2010 and rebuilding to robust 9. 6% growth and $11. 7 billion in revenue in 2012 § The troubled economy will affect HPC segments unevenly, with verticals such as automotive and financial services hit harder than oil and gas, or government and academia § The HPC storage market will remain strong through the recession

Conclusions • Standards-based clusters will gain market share in the price-sensitive economy, but more HPC sites will experience retrograde performance on some codes • More petaflop systems will arrive in 2009, and “easeof-everything” solutions will proliferate at the low end and beyond • HPC supply chain use will start to become a metric for industrial competitiveness • Because of its mission-critical nature, HPC will exit the recession the way it entered it as a bright spot in the IT space

Questions? Please email us at: ejoseph@idc. com sconway@idc. com jwu@idc. com rwalsh@idc. com

Why Is Commodity Hot? . . Price! HPC All Servers Processor Summary, 2008 Ave. CPUs / ASP ($K) System $K / CPUs / $M x 86 50. 7 21 2. 4 419 RISC 98. 3 10 9. 5 105 EPIC 55. 9 7 7. 6 131 873. 4 12 73. 6 14 Vector

From Our New End-user Study