
Transcription
WHITE P APERThe Business Value of Amazon Web Services AcceleratesOver TimeSponsored by: AmazonRandy PerryStephen D. HendrickGlobal Headquarters: 5 Speen Street Framingham, MA 01701 USAP.508.872.8200F.508.935.4015www.idc.comDecember 2013EXECUTIVE SUMMARYIn early 2012, IDC interviewed 11 organizations that deployed applications onAmazon cloud infrastructure services. The purpose of the IDC analysis was tounderstand the economic impact of Amazon cloud infrastructure services over time,beyond the well-documented benefits of reduction in capex and opex. Specifically,IDC set out to understand the long-term economicBusiness Value Highlights:implications of moving workloads onto AmazonApplications Running on AWScloud infrastructure services, the impact of movingapplications on developer productivity and Five-year ROI: 626%business agility, and the new opportunities thatbusinesses could address by moving resources Payback period: 7.1 monthsonto Amazon cloud infrastructure services. Theorganizations interviewed ranged from small and Software development productivity increase: 507%medium-sized companies to companies with as Average savings per application: 518,990many as 160,000 employees. Organizations in ourstudy had been Amazon Web Services (AWS) Downtime reduction: 72%customers for as few as seven months to as manyas 5.3 years. Our interviews were designed to elicit IT productivity increase: 52%both quantifiable information and anecdotes so Five-year TCO savings: 72%that IDC could interpret the full return-oninvestment (ROI) impact of Amazon cloudinfrastructure services on these organizations. The study represents a broad range ofexperiences, with companies discussing applications ranging from a small internallydeveloped application to a large commercial application with over 20 millioncustomers. The use cases reviewed offered a variety of steady-state and variablestate workloads.From these interviews, IDC was able to measure the impact of Amazon cloudinfrastructure services.Overall, the organizations interviewed recognized annual financial benefits averagingover 518,000 per application. The most significant benefit comes from movingapplications onto AWS infrastructure due to lower capital and operational costs. Thisreduction in capex and opex accounted for over 50% of the overall benefits found inthe study. IDC observed significantly increased developer productivity on Amazoncloud infrastructure services compared with prior implementations. The companiesinterviewed experienced greater developer productivity across all of the key softwaredevelopment life-cycle activities, which was a direct result of the extensive
development and runtime services that are provided by Amazon cloud infrastructureservices. Developer and IT staff productivity accounted for nearly 30% of overallfinancial benefits. The remaining benefits were driven by the flexibility and agility ofAmazon cloud infrastructure services, which make it easier to trial new businessmodels, support revenue-generating applications, and provide more reliable servicesto end users. These other benefits included: Benefits increase over time. There is a definite correlation between the length oftime customers have been using Amazon cloud services infrastructure and theirreturns. At 36 months, the organizations are realizing 3.50 in benefits for every 1.00 invested in AWS; at 60 months, they are realizing 8.40 for every 1.00invested. This relationship between length of time using Amazon cloudinfrastructure services and the customers' accelerating returns is due tocustomers leveraging the more optimized environment to generate moreapplications along a learning curve. The five-year total cost of ownership (TCO) of developing, deploying, andmanaging critical applications in Amazon cloud infrastructure represents a 72%savings compared with deploying the same resources on-premises or in hostedenvironments. The findings showed a 626% ROI over five years. End users benefited from fewer service disruptions and quicker recovery onAmazon cloud infrastructure services, reducing downtime by 72% and improvingapplication availability by an average of 3.9 hours per user per year. IT staff productivity increased by 52%. IT staff are thus able to improve support ofmission-critical operations. Amazon cloud infrastructure services had significantimpact on application development and deployment, reducing overall developerhours by 80%.The five-year ROI analysis shows that on average, the companies saw a paybackperiod of seven months and realized a five-year ROI of 626%.2014 UPDATESince this study was conducted in early 2012, AWS has introduced price reductionsnearly 20 times across Amazon EC2 and Amazon S3. IDC estimated what the impactof AWS's fee restructuring would be on the organizations that participated in the 2012study and determined that the overall fees would drop by 21% lowering the five yearTCO from 909,000 to 846,000.SITUATION OVERVIEWIntroductionFor a company that was founded in 1994 and began delivering cloud computing in2006, Amazon is a relative newcomer as a Fortune 100 company (Amazon.com firstappeared in the Fortune 100 in 2010), and AWS is one of the oldest cloud computingproviders in the industry (AWS started offering services in March 2006). Althoughentry into cloud computing was not an initial design point for Amazon, the demands of2#236029 2013 IDC
creating a reliable, secure, and scalable ecommerce presence, along with the desireto lower prices for its retail customers, led Amazon to focus on driving costs out of itsIT infrastructure. Amazon's focus on creating a service-oriented architecture put thecompany on the as-a-service fast track.Amazon's current services in support of application hosting, application management,security, data management, relational databases, nonrelational databases, payments,billing, storage, networking, content delivery, development, deployment, and workflowall come under the heading of AWS. This breadth and depth of AWS has enabledAWS to become the leader in cloud computing. While many enterprises initiallythought of AWS as an infrastructure services provider, this perception has expandedin recent years in light of the array of runtime services that AWS provides as aplatform for application deployment.IT Challenges TodayThe challenges and opportunities facing CIOs today have never been greater. Asenterprises become more reliant on IT to improve efficiency while simultaneouslydifferentiating services, IT has become a strategic asset in support of business,marketing, product development, and operations. This section identifies a number ofthe most pressing concerns that IT is tasked with addressing today. Confirmation ofthese challenges is readily available across the industry, and the issues have beenwidely reported on by IDC and many of the leading ISVs. Better alignment of IT with the needs of the business. IT has never beenbetter positioned to address process automation and process improvementneeds. Modern IT tools, techniques, and infrastructure are more effective thanever at supporting the needs of the business. Focus on core business processes. The opportunities for outsourcing,offshoring, and application hosting now provide enterprises with many optionsthat allow them to concentrate on improving support of mission-criticaloperations. With maintenance of existing applications accounting forapproximately 50% of IT resources, businesses must make hard decisionsregarding where to spend their time. Simplify, integrate, and automate. Both IT demands and complexity aremushrooming. Enterprises must constantly look for ways to simplify andrationalize their approach to IT through the use of more highly abstracteddevelopment and deployment tools and policy- and configuration-based services. Grow organizational profitability. Although IT is not normally considered aprofit center, it can contribute to profitability by enabling new business modelsand finding better ways to manage expenses. Standardize and consolidate IT assets. Consolidation has been on the ITagenda for years, and virtualization has helped decrease server sprawl whilesimultaneously increasing utilization. Standardization is especially importantwhere lower-level architectural decisions, such as networking and authentication, 2013 IDC#2360293
are concerned, but it is rapidly moving up the stack into areas such asmessaging and service enablement.Key Cloud Computing TrendsThe market for worldwide public IT cloud services encompasses packagedapplication software, platforms, and infrastructure that adhere to eight specific criteria— identified by IDC — that characterize a cloud service.Software as a service (SaaS) includes collaborative applications (such as messaging,conferencing, and team collaboration software) and business applications (such asCRM, ERP, financial, HCM, PLM, and SCM) delivered via the cloud services model.Revenue in the SaaS market was about 24.1 billion in 2012 and is expected toincrease at a compound annual growth rate (CAGR) of 20.8% through 2017Platform as a service (PaaS) includes application development and deployment toolssuch as application development software, application life-cycle managementsoftware, enterprise mashup and portal software, information management and dataintegration software, and middleware and business process management softwaredelivered via the cloud services model. Revenue in the PaaS market was about 3.8 billion in 2012 and is expected to increase at a CAGR of 29.7% through 2017.Infrastructure as a service (IaaS) broadly includes compute resources, storageresources, and system infrastructure software delivered via the cloud services model.Revenue in the Public Cloud IaaS market was about 9.3 billion in 2012 and isexpected to increase at a CAGR of 27.2% through 2017.Figure 1 provides a graphical view of the worldwide public IT cloud services marketsegmented by primary market. Vendor revenues associated with IaaS and PaaSaccounted for a 35% share of the overall market in 2012, which speaks to theimmense value that organizations place on developing and deploying applications onpublic infrastructure.4#236029 2013 IDC
FIGURE 1( Billions)Worldwide Public IT Cloud Services Segmented by PrimaryMarket, 017IaaSSource: IDC, 2013AMAZON WEB SERVICESDuring the late 1990s and early 2000s, as Amazon was emerging as the world'sleading ecommerce company, internal business requirements necessitated thatAmazon build out an application infrastructure that would support massive scale andreliability in the following areas: compute, parallel processing, storage, contentmanagement, data management (relational and nonrelational databases), transactionprocessing, messaging, queuing, payments, security, monitoring, and management.Competing IT objectives involving scalability and cost steered Amazon down the pathof service orientation. The services created during this IT transformation processultimately laid the foundation for AWS. Figure 2 identifies the key services providedby AWS. 2013 IDC#2360295
FIGURE 2Amazon Web ServicesSource: Amazon, 2013Amazon's focus on driving costs out of the company's large-scale ecommerceoperations led Amazon to move toward service orientation and exposing all resourcesas scalable and consumable services. This movement also ensured that thedevelopment culture at Amazon would be aligned with modern developmenttechniques and result in a platform that was flexible, agile, and extensible. This hasresulted in a high degree of AWS API envy across the industry. From an AWScustomer standpoint, this means it is easier to develop applications using AWS,easier to deploy preexisting applications on AWS, and easier to establish hybridoperations spanning AWS and private datacenters.6#236029 2013 IDC
BUSINESS V ALUEStudy DemographicsIn early 2012, IDC interviewed 11 organizations that had deployed AWS. Theorganizations ranged from small and medium-sized companies to larger companieswith as many as 160,000 employees. The organizations are based in North America,Europe, and Asia/Pacific and include representatives from real estate, banking,technology, media, sports, security, and management services industries. Theinterviews were designed to elicit both quantifiable information and anecdotes so thatIDC could interpret the full impact of AWS on these organizations. Table 1summarizes the demographics of the study.TABLE 1Study DemographicsCategoryAverageEmployees21,445Virtual servers from AWS1,232Applications being implemented with AWS9.7Criticality rating of applications deployed onAWS — from 1 (low) to 5 (high)4.5End users of IT services — internal1,372End users of IT services — external applicationusers2,230,248IndustriesReal estate, banking, technology, media,management services, security, sportsGeographiesNorth America, Europe, Asia/PacificSource: IDC, 2013The study base represents a broad range of experiences, with companies discussingapplications ranging from a small internally developed application supporting eightusers to a large commercial application with over 20 million customers. About half ofthe 11 companies are using AWS to deliver high-performance computingapplications, one-third are supporting Web applications, and one-quarter have bigdata applications. The use cases reviewed offered a variety of steady-state andvariable-state workloads. 2013 IDC#2360297
FINANCI AL BENEFITS AN ALYSISBenefits SummaryThe organizations in the study selected AWS to develop, deploy, and manage theircritical applications primarily because AWS offered the best scalability, time tomarket, and price. From the interviews, IDC was able to measure the impact of AWS.Overall, the organizations recognized annual financial benefits totaling over 5 million(over 518,000 per application) from the following areas: Reduced IT infrastructure and services costs. The most significant benefitcomes from rehosting applications on AWS infrastructure due to lower capital andoperational costs. This reduction in capex and opex accounted for a savings ofnearly 276,000 per application per year. Companies were able to consolidate,integrate, and standardize their infrastructure. Optimized IT staff productivity. By accelerating the application development anddeployment process, automating application management, and switching to IaaS,IT staff are now 52% more productive, saving nearly 150,000 per application peryear. IT staff are thus able to improve support of mission-critical operations. Enhanced end-user productivity. End users benefited from fewer servicedisruptions and quicker recovery, reducing downtime by 72% and saving nearly 32,600 per application per year. Increased business benefit. Many of the companies are employing AWS toenable new business models and support revenue-generating applications andwere able to increase annual revenue by over 1 million, which translates to morethan 64,000 in annual operating income.Figure 3 provides an aggregate view of these benefits.8#236029 2013 IDC
FIGURE 3Average Annual Benefits of Deploying Applications on AWS per Application600,000Total 518,990500,000 32,566 64,261( )400,000 146,355300,000200,000 275,807100,0000User productivityOperating incomeIT staf f productivityCost reduction: inf rastructure and servicesSource: IDC, 2013Cost ReductionCompanies today look to the cloud as a way to shift expenses from capital tooperational budgets. All of the companies in the study not only shifted their budgetsbut also significantly reduced their costs and time to market for delivering thesecritical applications. In terms of annual budgets, they were able to replace 3.66 incapital costs with 1.00 in new operational costs, namely Amazon EC2 costs.In this study, we assessed IT infrastructure savings resulting from migration to the cloudin the areas of server, storage, networking hardware and software, and server hostingservices. On average, each company was able to reduce its annual infrastructure andservices costs by nearly 276,000 per application in the following areas, as shown inFigure 4, as a result of the shift from capital costs to operational costs. Servers (hardware and related software) — includes the annual capex for thephysical servers replaced by AWS (average 400 servers replaced per customer)as well as the operating system and related software efficiencies realized bymoving to the cloud environment (security, management, etc.) Hosting and other services — includes savings in the areas of collocationservices (four of the companies had collocation services) and/or managedservices as a result of moving to the same level of services from AWS Networking equipment and bandwidth — includes WAN equipment andsavings from more efficient use of bandwidth 2013 IDC#2360299
Storage — includes savings from moving storage to the cloud, resulting in moreefficient utilization and redundancy of storage assets and a 25% reduction in theoverall storage cost per terabyte (TB) in useFIGURE 4Annual Infrastructure and Services Cost Benefits of AWS per Application300,000Total 275,807 13,626 21,114250,000 50,409( )200,000150,000100,000 190,65850,0000StorageNetworking equipment and bandwidthHosting and other servicesServersSource: IDC, 2013IT Staff EfficiencyOur analysis focused on tracking a critical application through the costs associatedwith development, deployment, ongoing management, infrastructure, and staff tosupport the infrastructure. As discussed earlier, AWS had a significant impact onapplication development and deployment, reducing overall developer hours by 80%.This increased productivity of developers means that organizations can deliverapplications to market faster as well as increase the volume of applications beingdeveloped.10#236029 2013 IDC
Development and Deployment EfficiencyAn important area of consideration for this study was developer efficiency across thesoftware development life cycle (SDLC). In our conversations with AWS customersfor this study and the projects we evaluated, we found mostly hybrid approaches tothe SDLC where tasks were split across in-house systems and AWS. Although somecustomers performed all aspects of their development project on AWS, this was notthe norm and is explained by sunk costs on the part of customers, inertia, and cautiondue to the new technology and business models of cloud services.Figure 5 shows the improvements that we observed in developer efficiency across theSDLC. These findings should be interpreted as follows: A 500% improvement inefficiency means that developers could perform tasks five times faster with AWS thanwith other in-house alternatives. A 200% improvement in efficiency means that AWSenables developers to perform tasks twice as fast as other in-house alternatives.FIGURE 5Comparison of Developer Efficiency with AWS andIn-House Alternatives(Improved ef f rationDeploymentTotalNote: A 500% improvement in efficiency means that developers could perform tasks five times faster withAWS than with other in-house alternatives.Source: IDC, 2013The high scores received by AWS on development and testing were due to theavailability, scope, and depth of AWS APIs. AWS APIs enabled developers to moreeffectively utilize services during application development, which reduced time spentwriting additional custom code. The maturity of AWS also means that its services arewell vetted and therefore lead to higher-quality applications with reduced defects andless time spent resolving defects. While many platform vendors will claim that suchbenefits stem simply from the adoption of platform technologies, we believe that thesedeveloper benefits are closely tied to the capabilities of the cloud infrastructure. For 2013 IDC#23602911
example, the automated provisioning, dynamic scalability, management, andmonitoring that are part of AWS mean that the development of applications can besimplified because many key services that govern application behavior no longerhave to be coded; instead, they can simply be configured as part of deployment. Thisapproach also helps explain why deployment efficiency is commensurate withdevelopment and testing efficiency.12#236029 2013 IDC
Integration efficiency received a lower score than other life-cycle tasks but still wastwice as efficient as alternative approaches. Integration involves legacy assets andpotentially hybrid datacenter interoperability, making integration challenging. Whilethe improvements in integration were still significant, these gains were not as large asthose for other SDLC activities, reflecting both the complexity of opening up andlinking to applications not originally engineered for the Web and the low incidence ofintegration tasks that we were able to measure.Post-Deployment Staff ProductivityAs with most technology, if you make something easier, quicker, and cheaper to do,you will do more of it; so it goes, especially with application development. Becausecompanies in the study could create applications with fewer staff hours, they tendedto deliver more frequent upgrades or reallocate resources to create new applicationsthat they had not previously planned. The net result was that they were able togenerate more applications yet reduce the costs to develop and deploy by 25%.Post-deployment IT staff productivity savings result primarily from the reducedinfrastructure footprint, requiring less maintenance and support. The combined costsfor server, storage, and network support are reduced by 68%, which includes the costavoidance from not having to increase staff to meet new business requirements. Inaddition, AWS management services reduced application management costs byanother 70%. Figure 6 shows the annual post-deployment IT staff productivitybenefits per application in terms of FTEs. 2013 IDC#23602913
FIGURE 6Annual Post-Deployment IT Staff Productivity Benefits of AWSper .530.0OtherAWSManagement savingsDevelopersServer managementApplication managementStorage managementNetwork managementSource: IDC, 2013Business BenefitsThe most significant impact on the business operations of the organizations weinterviewed was the enhanced agility delivered by AWS. This time-to-market benefit isoften cited as the key reason for adopting cloud services. Seven of eleven companiesreported significant improvements in agility. Not all contributed to revenue. Four of14#236029 2013 IDC
these organizations were able to discuss how revenue-related agility improvementsenabled new business models: "We can get transactions quicker in our accounting software now. Our people aremore productive. It's really seamless to the end user in that regard. They physicallyhave more time to do that work. Estimated additional annual revenue is 500,000 ." "With AWS, we are able to launch some of the services instantaneously. It wouldtake many months, if not a year or more, to build out that whole infrastructurefrom scratch. Estimated additional annual revenue is 2 million ." "In half a year that we've had this service on market it's become almost our mostpopular service. I mean we're talking about millions of dollars in the last six months.Low millions so far. But we think that it's going to generate hundreds of millionsover the life of this project. Estimated additional annual revenue is 3 million ." "Our ability to go to market with our new program. We would not have been ableto get that out. Estimated additional annual revenue is 200,000 to 500,000."User ProductivityMany companies opt for IaaS on an application-by-application basis and usually whenfaced with a new requirement. As businesses become more demanding and thestakes get higher, CIOs are concerned that their current resources cannot deliver onthe agility, scalability, or robustness that the new application requires. The robustnessor quality of service is often measured in unplanned downtime and is the number onereason companies switch providers. In this study, 80% of the organizations reportedthat Amazon EC2 had reduced downtime and that they were able to reduce theirdowntime by 72% per user, saving roughly four hours per user per year. In otherwords, the companies interviewed were able to improve availability from 99.495% intheir on-premises/hosted implementations to 99.975% on AWS, increasingapplication availability by an average of 234 hours per user per year. Table 2summarizes productivity improvements achieved with AWS.TABLE 2User Productivity — Performance KPIs (as Reported byCustomers)Downtime events per yearHours per yearDowntime hours per user per yearAvailability (%)OtherAWS% .97595Source: IDC, 2013 2013 IDC#23602915
Total Cost ComparisonBy moving from on-premises or mixed on-premises and hosted solutions to AWS cloudservices, organizations in our study were able to reduce the TCO for the averageapplication by 72% over a five-year life cycle. TCO savings include reducingdevelopment and deployment costs (by 80%), application management costs (by 52%),and infrastructure support costs (by 56%) and replacing 1.6 million in infrastructurecosts with 302,000 in AWS costs. The five-year TCO is shown in Figure 7.2014 UPDATESince our last study in early 2012, Amazon EC2 and Amazon S3 haves introduced pricereductions nearly 20 times across certain key instance types, sizes, and configurations,in some cases significantly lowering overall customer spend. For example, AmazonEC2 Reserved Instances (RI) is a mechanism by which one-time fees are paid, andusers in turn receive a significant discount on the hourly usage charge for that instance(AWS says up to 71% lower than on-demand instances). AWS recently dropped theprice of most Linux/Unix RI instance families by 27%. AWS has broadened what it callsits "free usage tier" - a starter tier for AWS beginners/trailers - to include most versionsof Windows server, a very common configuration choice for proof-of-concept (POC)customers, as well as most flavors of Linux. AWS lowered prices in early 2012 forseveral instance sizes of standard storage, Relational Database Service (RDS), andElastiCache, and followed this action up four more times in 2012 and 2013 with lowerprices for most Amazon RDS. In addition, a key requirement for most users is multiavailability zone (AZ) for better availability and performance of transactional apps, andmulti-node (for backup and redundancy), and AWS has cut its multi-AZ charges tocustomers twice in the past 18 months, by an average of 19%. Prices for AWS SimpleNetworking Service (SNS), and Simple Workflow Service (SWS) have already beenlowered once since 2012.AWS's express reason for lowering cost is that its scale and unique industry weightallow it to leverage OEM/ODM relationships for lower cost, custom-design software andhardware and utilities management systems for optimal consumption, and otherwiseinnovate and apply its first-mover learnings to pass savings along to customers.IDC estimated that if the 11 AWS customers we had interviewed in 2012 to support thisstudy were to migrate to AWS today, they would experience their overall fees droppingby an additional 21%; lowering the five year TCO from 909,000 to 846,000. Itsimportant to note that the other business value results reported in this whitepaper do nottake into account these new lower prices from AWS due to the timing of our AWScustomer research.16#236029 2013 IDC
FIGURE 7Five-Year TCO of AWS per Application3,500,0003,000,000( AWS (pricing as of AWS (pricing as ofJuly 2012)November 2013)DevelopmentDeploymentApplication managementInf rastructure managementInf rastructureAWS f eesSource: IDC, 2013Benefits Increase Over TimeOrganizations in our study had been AWS customers for as few as seven months toas many as 5.3 years. While all customers had enjoyed positive returns oninvestment, there is a definite correlation between the length of time they have beenAWS customers and their returns. This relationship between length of time usingAWS and return is due to their leveraging the more optimized environment togenerate more applications along a learning curve. At 36 months, the organizationsare realizing 3.50 in benefits for every 1.00 invested in AWS; at 60 months, theyare realizing 8.40 for every 1.00 invested. 2013 IDC#23602917
ROI AN ALYSISIDC uses a discounted cash flow methodology to calculate the ROI and paybackperiod. ROI is the ratio of the net present value (NPV) and discounted investment.Payback period is the point at which cumulative benefits equal the initial investment.IDC uses the NPV of the savings and increased revenue over five years in calculatingthe ROI and payback period for the deployment. The NPV of the savings isdetermined by subtracting the amount that would have been earned by investing theoriginal sum in an instrument yielding a 12% return (to allow for the missedopportunity cost that could have been realized using that capital).Table 3 presents IDC's ROI analysis for the deployment of AWS to replace onpremises or other hosting services. This ROI analysis constitutes a five-year view ofthe financial impact of AWS on a per-application basis. A detailed outline of IDC'sstandard ROI methodology is provided in the Appendix of this document.The five-year ROI analysis shows that on average, the organizations in this studyspent 286,000 per application on AWS and received 2.1 million per application inbenefits for an NPV of 1.8 million. The companies saw a payback period of sevenmonths and an ROI of 626%.TABLE 3Five-Year ROI AnalysisBenefit (discounted) 2,078,626Investment (discounted) 286,357Net present value (NPV) 1,
Amazon's current services in support of application hosting, application management, security, data management, relational databases, nonrelational databases, payments, billing, storage, networking, content delivery, development, deployment, and workflow all come under the heading of AWS. T