OGF20 May The 451 Group Grids: the means to what ends? Steve Wallage Director of Research William Fellows Principal Analyst
OGF20 May The 451 Group Technology industry analyst company Agenda: enterprise IT innovation Analysis that is timely; insight that is dynamic Network of 700+ client organizations globally Vendors Investors Bankers Early-adopters
OGF20 May Grid Adoption Research Service (GARS) Tracking 250+ enterprise early adopters Across a range of sectors: Financial services, telco, pharma, film and gaming, manufacturing, energy, hi-tech, healthcare Common business and technology functions: Measuring value, software licensing, data management, utility computing, application development/design, state of market Enterprise Computing Strategy industry summits
OGF20 May EARN-IT: Early Adoption Research Network - IT GARS - Grid Adoption Research Service CAOS - Commercial Adoption of Open Source Security, mobile enterprise…
OGF20 May Our work with the EC FP6 GridEcon project looking at economic and business models for grid usage Leading technical working group (TWG7) on business models and SLA for grids European Grid Co-ordination Technical Committee Member of NESSI FP7……
OGF20 May view of market - agenda Overall state of the market Drivers Challenges Vertical differences
OGF20 May Levels of deployment
OGF20 May What to call it? 70% of respondents said there is a better term than grid to describe their distributed computing architecture:
OGF20 May Drivers – GARS users 77% 57% 49% 41% 26% Time to Market Competition Do New Things Save Money Improved Performance
OGF20 May Drivers – a broader look
OGF20 May Drivers – proving the business case Tangible vs intangible Buy-in Proof of concept Competitive advantage
OGF20 May Challenges – GARS interviews 46% 33% 31% 28% 20% 19% 18% 15% 14% 1% Semantic Web Skills Shortage App Dev Lack of Standards Prove ROI Grid to Utility Grid to SOA Grid Enabling Apps Security Data Mgmt Cultural Bandwidth Licensing
OGF20 May Challenges – a broader look
OGF20 May Challenges – experiences from a different perspective Sensitive data, sensitive apps (medical patient records) Different organizations get different benefits Accounting, who pays for what (sharing!) Security policies: consistent and enforced across the grid Lack of standards prevent interoperability of components Current IT culture is not predisposed to sharing resources Not all applications are grid-ready or grid-enabled Open source is not equal open source (read the small print) SLAs based on open source (liability?) Static licensing model dont embrace grid Protection of intellectual property Legal and tax issues (FDA, HIPAA, multi-country grids)
OGF20 May Cultural challenges Ownership and control CxO buy-in Importance of stakeholders Internal SLAs
OGF20 May Deployment – by vertical
OGF20 May Vertical market - differences Software licensing Range of applications Use of desktop scavenging Use of outsourcing Use of open source
OGF20 May Where are we going? William Fellows
OGF20 May Levels of deployment
OGF20 May Next - support broader IT objectives
OGF20 May What do we know?
OGF20 May The sweet spots of activity are: Enterprise utilities Platform for shared services In other words, a technology infrastructure and an application infrastructure.
OGF20 May The key technology drivers Grid technologies Virtualization SOA, SOLE, SOIT, SOE…. New approaches to power, performance, space
OGF20 May Virtualization Enterprises are experiencing benefits from infrastructure virtualization Virtualization is being applied at every layer of the IT stack - compute, network, data and apps Concerns: sweeping complexity under the carpet, performance overheads, license costs, vendor lock-in, too many moving parts? Virtual appliances rival SaaS? Do grid and virtualization intersect, converge or collide?
OGF20 May One approach: a layered view (from EGA reference model) Business process / service Reference DataRisk ManagementCustomer Portal Virtualized PlatformData GridCompute GridServer Farm Platform InstanceDatabaseApp ServerWeb Server Virtualized Operating Environment NFS, CIFS, SMB, NAS Virtual Machine Monitors, Solaris Containers, BSD Jails Load balancing, Global IP, Virtual IP Operating Environment File systems e.g. NTFS, Ext3 Operating Systems e.g. Linux, Windows Network protocols e.g. TCP/IP, UDP Virtualized PhysicalLUNsHypervisorsVLANs PhysicalDisks, Array Controller, SAN switches etc. Servers, Blades etc. Switches, Routers etc. StorageComputeNetwork Each physical layer provides abstraction to the layer above Each virtualized layer provides a flexible mapping/management
OGF20 May Grid and virtualization: vectors intersect Virtualization - a vehicle to realize the dream of grid computing? Mapping virtual workspaces to physical resources - virtual grids The end of grid? Battle for control - vendors extending into other territories
OGF20 May Grid and enterprise utility Leading adopters implementing grid economy across shared resources Requires willingness to participate in a shared IT infrastructure Self-service is a driver SaaS and open source are cost-effective and additive Next steps - SLA monitoring, policy management, chargeback across heterogeneous resources
OGF20 May Grid and public utility computing Grid delivers economies of scale for next-generation hosting providers = increased adoption Are we ready for utility computing now? (Yes, but don't call it that) So what is interesting? Issues: multi-tenanted, per-drink pricing, software licenses, SaaS pricing complexity, lack of app services, loss of budget predictability
OGF20 May Grid and public utility computing Users with enterprise utilities are more likely to examine other outsourcing options Back in the conversation but wont be an 'all or nothing' play A lot of marketing, but not yet a market Telcos become IT services providers. Google, Amazon, eBay take on incumbents (eBay financial markets?)
OGF20 May IT as a utility 80% culture change - 20% technology integration Back to the future? Charge to a business metric not CPU/hour - at the service delivered to the user not resources used at backend Cheaper in-house? Outsourced CPU suppliers need better pricing models
OGF20 May Rebirth of cool Enterprise datacenter sales used to be two dimensional - what is the performance and what is the cost? Power consumption, heat dissipation, space and their SLAs are top of procurement conversations Blades = less space but more more heat Rebirth of mainframe cooling techniques - IBM cool door, HP cool fan, Green Grid.. Other approaches: pooling/outsourcing for peak loads, multicore, FPGA, cell, GPUs, streaming CPUs
OGF20 May Rebirth of cool US EPA Energy Star program: every dollar spent on IT equipment, $3 to $4 is spent on operating it over its life Energy costs for UK businesses have increased by 57% during the last 12 months, and now form a significant element of operational expenses, often greater than IT equipment depreciation and sometimes greater than real- estate costs How to factor into ROI? Return on Environment - new metric for technology innovation 451 Group report: EcoEfficient IT
OGF20 May Banks verdict: grid is good for… Looks capable of driving sustainable, long-term and linear cost savings and performance improvements Use of grids to scale across the organization and support additional activities such as exotics Can common approaches (above the waterline) drive differentiation and competitive advantage? Return to spending on innovation instead of managing complexity Utility, flexibility, scale - agility Scale: This year's exotic can easily become next year's flow.
OGF20 May Who do we know about?
OGF20 May Data management Problem Interface between compute grid and data grid/cache File systems are a bottleneck – storage startups solve it? Storage not kept up with compute power Latency is a business issue not a technology issue - for faster business cycles the time to execute must be less than the time to act Innovation Storage/file systems (EMC, NetApp, Ibrix, Nirvana, Acopia, NeoPath, Isilon) database (Xkoto, Vertica), clustering (PolyServe, Panasas, Isilon, Oracle), caching (Tangosol, GemStone, GigaSpaces, Composite), data integration Google BigTable, MapReduce, Sawzall; Microsoft Dryad
OGF20 May Metering, billing, chargeback Problem Measuring virtualized compute, network and storage resources Allocation of resources and priority assessment Rating, mediation Policy enforcement Unit of measure Innovation Evident Software, LeCayla, Provment, Iontas, Itheon, IBM CIMS, Digital Fuel
OGF20 May Managing virtualization. Implementing SLA/policy Problem Managing VMs Automation of tasks/run book Enforcing SLAs/policies Performance analysis Semantics Innovation VMware, Xen, Scalent, DataSynapse EverGrid, ToutVirtual, VirtualIron, IBM Meiosys, Platform, UD, Trigence, Dunes OpsWare/iConclude, BladeLogic, Cassatt, Egenera, RealOps, Enigmatec, Opalis,Tideway, OpTier Dtrace, SystemTap, Sawzall
OGF20 May Applications Problem Run existing applications unchanged, self-service A platform for new app development or chopping up older applications? Levels 1 to 5… New approaches to multi-threading, parallelism Batch to interactive, real-time: mixed workloads Innovation Grid or fabric-aware application servers provide virtualized containers for apps and app servers IBM XD, UD/SAP, Oracle, DataSynapse, Platform, BEA, Scalent, Appistry, Aspeed, 3Tera, Trigence, EverGrid
OGF20 May Security Problem Neglected by many users, believe protected within their own firewalls Biggest threat is internal External challenges - malicious usage, stealing algorithms, using the grid against its owner, using the grid for unauthorized programs Trusting of resource brokers, other internal (can you trust another department ?) and external data sources The 451 Group has formed a panel to develop security recommendations for organizations deploying grids
OGF20 May Where are we going?
OGF20 May grids not The Grid Resurgence of grid HPC or NGDC? Convergence - datacenter automation Could use some some open standards - but waiting is too risky (getting left behind) Requires maturity of processes Some of todays tools are sufficient - but we are in the inventory period in terms of usage Business and senior management expectations exceed capabilities?
OGF20 May Beyond using Grid to solve HPC Define new workloads for grids Platform to support SoA Create shared service architecture Enable collaboration Standardized semantics Build composite applications Create virtual organizations
OGF20 May grids - through 2010 Virtualization allows grid to be absorbed into to the enterprise fabric Convergence of grids, multicore and virtualization will have a profound effect Silos to horizontally integrated resources - enterprise utilities Downstream of HPC: mixed workloads, risk management in other verticals, analytics, MDM, ERP Batch and real-time/interactive blurring Mitigate risk through flexibility
OGF20 May grids - through 2010 Continued confusion between grid, SOA, virtualization and utility computing Vendor execution and partnerships - more important than vision Latency is a business issue not a technology issue - for faster business cycles the time to execute must be less than the time to act Users increasingly socialize their IT models Innovation: companies cant just cost-cut their way to prosperity The end of grid? No. Grid is the means to many ends
OGF20 May The 451 Group Grids: the means to what ends? Steve Wallage Director of Research William Fellows Principal Analyst
OGF20 May Thankyou. Questions?