IT Infrastructure Trends & Directions October 3rd 2002 Andrew Butler Vice President & Research Group Director research consulting measurement community news
Macro Trend on Application and Infrastructure 1980s 1990s 2000 to 2005 2006 to 2010 Big Iron Age Distributed and Internet Revolution Age of Maturation and Platform Shake-out Renaissance Era Central IT Decentralized Datacenter LOBs Standard hardware Standard applications stack Deep metrics High quality of service Policy-based management Platform choice Nonstandard app. stack Many suppliers Chaotic management environment; little instrumentation Low quality of service Less platform choice Move toward a few standard app. stacks Improved instrumentation Push for SLAs/metrics Process maturity NSM supplier shakeout Active management Effective root-cause analysis Self-healing application services Policy-based management Autoprovisioning Service views via autodiscovery
Bandwidth Becomes More Cost-Effective Than Computing Compute cheaper than bandwidth Terse data expanded on use GUI processing at client Local copies of data Consumer location matters Distributed centers Application on few, big servers Bandwidth cheaper than compute Full images stored Thin is in! Central copies of data Centralization for scale advantage Consolidated Application on many, little servers Grid computing based on P2P Advertisement Discovery Use Issues and inhibitors Quality of service Economics Privacy & security
Most Major New Systems Will Be Inter-organizational Conflicting priorities Incompatible NSM tools No common owner Not built for x-enterprise Must integrate NSM tools IT processes Help desk Must develop Coordination Policies Virtual teams
Continued Consolidation of Vendors in Many Segments Number of Software Vendors Half of today s software vendors gone by 2004 Big brand stacks slow down their innovation, products mature Most surviving small firms just complement the big brand stacks 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Innovators of Internet boom burn themselves & venture capital cash Oligopoly of fewer vendors Return of green shoots of innovation, new entrants and big VC cash
Business Units, Not IT, Will Make Most Application Decisions Resource Function Frontier/ Strategic Enhancement/ Efficiency 9% 30% 28% 33% The lines of business want to see more of a transition to innovation Frontier Strategic linkage Risk balancing Utility/ Maintenance 38% 0 10 20 30 40 50 60 Percentage of IT Resource Allocation 63% 70% Utility High reliability Lowest cost Future Today Enhancement ROI-based Prioritized
Where is the Money Going? EMEA IT Market Forecast, 2002-2003 Billions of $ 400 350 300 250 200 150 100 50 Percent Growth 1.80% 1.60% 1.40% 1.20% 1.00% 0.80% 0.60% 0.40% 0.20% 2002 2003 Growth Rate 0 Hardware Software IT Services Telecom Equipment Telecom Services 0.00% Source: Gartner Dataquest, 2002
Consolidation Platforms Rise, Midrange Collapse Billions of $ 2,500 EMEA Server Market: 1H02 +29% 2,000 +13% Linux Windows 1,500-13% -28% -53% Other UNIX 1,000 500 0 Server Appliance < $5k $5k-$10k $10k-$25k $25k- $100k $100K- $500K >$500K
Pressure on Microsoft From Unix Resilience & Linux Adaptability Billions of $ 25 Value of Shipments by OS 20 15 10 Windows Unix Linux Other 5 0 2002 2003 2004 2005 2006 2007 Source: Data Center Conference Survey (December 2001) and Dataquest Forecasts (1H02) Would you be willing to replace or bypass current Microsoft Windows servers with Linux if nearequivalent applications existed for Linux, even if the software is a commercially licensed package? Yes: 53% No: 47%
Matching Server Technology Choices to Enterprise Needs: 2002-2007 Processors 64+ 32 16 Traditional tower or frame-based servers Compute clusters G r i d 8 New NUMA 4 2 1 Monolithic computing (scale up) Rack dense Blades Distributed computing (scale out)
Linux in the Enterprise Client Devices Internet Access Line of Business Enterprise Data Infrastructure Server/Blade Business Application Server Data/ Content Server/ Warehouse Desktops Proxy Caching VPN Firewall WAP VOIP Gateway Network Edge Servers Directory Security Load balance File/print Web E-mail NAS ERP SCM CRM HR Databases Documents Images Compute Clusters R&D, biotech, geophysical, energy, environment, visualize
zseries and Linux: Where Are The Savings? Improved availability Reduced people costs If the number of images is reduced significantly or If administration efficiency is improved sufficiently After consolidation alternatives are ruled out Reduced software costs If engine-based or switching to open source Until software vendors adjust pricing Reduced hardware costs If the number of systems is reduced significantly If moving from a non-linux environment After consolidation alternatives are ruled out Efficient use of hardware If workload peaks are offset If over-provisioning is not required
Commercial ISV Support Profile Size of Application Portfolio IBM OS/400 Windows 2000 HP-UX Solaris/ IBM AIX SPARC IBM z/os Caldera Open UNIX HP Tru64 Unix HP OpenVMS Red Hat Linux Other Linux HP NonStop Linux on zseries Independent Software Vendor Enthusiasm Safe for virtually all mainstream deployment Safe for many target segments and installed base leverage Limited verticals, niche segments and installed base only May 2002
Match Consolidation Options To Required Benefits Application OS Circa 40 (1.6Mtr) Up To 300 (1.6Mtr) Hypervisor Frame Generic Rack Firmware/ Software Software Server/Node Rack Blades/ Bricks Physically Partitioned Logically Partitioned Workload Manager Footprint Manageability Vertical Scaling Horizontal Scaling CPU Utilization Isolation June 2002 Best Worst
Today s Computing Geography Static and Unsharable Islands Service A Service B Service C (And Many Others) Edge Application Database Current state is inflexible and does not reflect business priorities Inefficient Overprovisioned Hard to manage STRATEGIC PLANNING ASSUMPTION: Policy-based computing services are inevitable, rolling out in phases through 2010 (0.8 probability), driving lower IT costs, greater quality of service and greater agility. Those IT organizations and service providers not embracing them will risk survival, because their costs will be much higher than their competition.
Gartner s Server Evaluation Model -- Weighted Data 400 350 300 250 200 150 100 50 0 Technology Momentum Business Practice IBM zseries HP NonStop Sun 15K HP Superdome IBM Regatta FSC Primepower IBM x440 HP Proliant Dell PowerEdge Unisys ES7000
Storage Network Adoption Rate 100% File based Not general-purpose, niche Simple to deploy Public & Standardized 80% 60% 40% 20% 0% 1998 1999 2000 2001 2002 2003 2004 2005 Highest Performance Management overhead Proprietary SCSI over IP Block access Low cost, entry level SAN Interconnection over wide area Lower performance Block based Private Over-hyped No-standards yet NAS SAN iscsi SAN Fibre Channel Direct Attach
PC Vendors - Fighting More, Taking Less Cost Breakdown for a Midrange Desktop PC 100% 90 80 70 60 50 40 30 20 10 0 4Q98 4Q01 4Q04 Channel Margin Vendor Margin OS Other Processor
Software Tortoise, Hardware Hare Action Item: Organizations should begin the due diligence to determine whether a four-year replacement strategy is applicable. GHz MHz 3+ 1.8 900 Hardware capacity will continue to outpace software requirements for the foreseeable future P4 700 500 300 100 3.11 95 Need for speed Pentium NT4 PII PIII 98SE 2K Excess Capacity XP 94 95 96 97 98 99 00 01 02 03
Vendor Suitability Global Vendors g Multi-regional support g Large-scale integration Large Enterprise Gov t Education SMB g Local presence g High touch g Accountability g Specific business knowledge Regional Vendors
Infrastructure Recommendations Server consolidation is driving the demand for agnostic server fabrics and policy-based computing. We encourage clients to target servers three years or older for replacement and possible consolidation. Do not select the storage technology first and then look for solutions. Directattached storage (DAS), SAN and NAS are all viable solutions, but typically they are for different situations. Check your challenges, set your targets and then select the most appropriate technology. The launch of Itanium II alone will not stimulate a volume market; expect that widespread Windows and Linux Itanium adoption will not commence until at least 2004. ISV enthusiasm is still a major selection criterion for server platform selection. But the volatile market makes it hard to guess the ISV leaders of tomorrow, and how and when today s ISVs will endorse Linux and IPF.
IT Infrastructure Trends & Directions October 3rd 2002 Andrew Butler Vice President & Research Group Director research consulting measurement community news