Success in the cloud, why workload matters

  • View

  • Download

Embed Size (px)


Learn about how cloud computing has accelerated IBM’s ability to innovate and do so in ways that improve how IBM delivers services and support. In short, it has become a catalyst for business transformation at IBM.

Text of Success in the cloud, why workload matters

  • 1. IBM Office of the CIOMarch 2012Thought Leadership White PaperSuccess in the cloud:Why workload mattersObservations from IBMs own cloud transformation

2. 2Success in the cloud: Why workload mattersContentsWith the implementation of cloud computing internally, acrosssix fundamental IT workloadsdevelopment and test, analytics,2 Executive summary storage, collaboration, desktop and production application3 Introductionworkloadswe have witnessed striking improvements inefficiency while capturing some impressive savings in capital3 IBMs impetus for adopting cloudand operations. Consider that:3 The importance of workload selection IBM development teams have seen server provisioning and5 Clouds biggest impact at IBMconfiguration drop from five days or longer to as little as one hour. IBMs development and test cloud has virtually elimi-9 IBMs other successful workload migrations nated IBMs testing backlog, speeding new development and11 Clouds role in reinventing IBMs businessenabling applications to reach the market sooner. IBMs analytics cloud put an end to siloed business intelligence11 Conclusion(BI) and the six-figure funding required for new BI projects.12 For more informationOrganizations across IBM are tapping into a centralized analytics cloud for tools and intelligence aggregated from hundreds of information warehouses. The associated savingsExecutive summaryare expected to reach tens of millions over five years.Like other performance-driven companies, IBM is continuously IBMs block storage cloud cut the cost-per-byte of data storedchallenged by stakeholders to drive new revenue opportunitiesby nearly 50 percent at one of the first IBM facilities in whichand efficiencies while lowering costs. As such, cloud computingit was implemented. This has allowed the facility to accommo-with its widely-touted benefits made a convincing case for adop- date the explosive growth in storage demandupwards oftion. And the technology model has lived up to expectations. By25 percent annuallywithout increasing its total storageproviding a platform to standardize and automate key businessbudget, and it is expected to do so for four straight years.applications, cloud has enabled dramatic reductions in IBMs ITsupport costs and major improvements in workplace efficiencyThrough these and other internal cloud implementations,and resource use. More than that, cloud computing has accelerat-IBM is seeing firsthand the impact that cloud can have on theed IBMs ability to innovate and do so in ways that improve how business. There is little question about cloud computings poten-we deliver services and support. In short, it has become a catalyst tial to drive efficiency and lower costs. But the workloads a com-for business transformation at IBM. pany selects for migration to the cloud and their affinity for theattributes inherent in the cloud model have a lot to do withclouds success as a transformative tool for the business. 3. IBM Office of the CIO 3Introduction IBMs impetus for adopting cloudIBMs IT leaders viewed cloud computing as an opportunity to IBMs intent in adopting cloud was not unlike any otherradically simplify aspects of an IT operation that had grown business. With its sizeable prospects for controlling IT costs,complex and less productive. Based on our own cloud research accelerating new capabilities and delivering as a service every-and years experimenting with similar technology models, we thing from infrastructure resources to business processes, cloudunderstood the potential. Cloud could help IBMs development had the potential to radically change the economics of IT. Itand testing teams provision their own server and storage capac-could drive up the value that IT provides to the business whileity without week-long delays or involvement from systemlowering capital and operational costs. Cloud also offered theadministrators. Cloud could help us move employees from ameans to deliver on the increasing demands of IBM employees,resource-straining traditional desktop environment to a virtualbusiness partners and customers who have come to expect a newdesktop environment, facilitating new deployments, upgradesstandard of service.and end-user support. Cloud could facilitate online collaborationamong IBMs global workforce, making it easier for employees More importantly, these cloud-enabled cost and operationalto interact, share ideas and innovate with clients, business efficiencies had the potential to transform IBMs business, notpartners and each other. just in how IT resources and services were delivered, but in how IBM conducted business with clients and partners around theCloud has succeeded in doing all of these things, but much has globe. Cloud provided the innovative and collaborative platformbeen learned along the way. As with any new technology, cloudand the computing flexibility to reinvent business at best deployed in the right circumstances and with the rightworkloads. One of the most significant determinants of success The importance of workload selectionin moving to the cloud is the careful selection of those work- Cloud computing has captured the attention of technology andloads. Some workloads are simply a better match for cloudbusiness leaders alike, but the actual value it delivers to thecomputing, with more to gain from clouds intrinsic features.organization varies with the application. When applied to the right workloads, cloud can deliver game-changing value. WhenThis paper shares IBMs observations and recommendations applied to the wrong workloads, clouds value over traditionalwith respect to workload affinity for cloud computing. Itdelivery models can be diminished or lost altogether.describes the quantitative and qualitative value that IBM hasachieved with six common IT workloads. And while each of In selecting workloads for cloud, it is important to consider thethese workloads has benefitted from the cloud, they have soft, intangible benefits as well as the more visible and easilybenefitted in different ways and to different degrees. Three quantifiable benefits. After all, intangibles like customer satisfac-workloads, development and test, analytics, and storage have tion and quality of service are often significant elements of cloudproduced the most stunning results to date. They have trans- value. For some workloads, they may represent the lions shareformed the way whole groups of IBM users do their jobs,of the value delivered.enabling them to speed new development and uncover newsources of revenue, among other things. The results achieved Clearly, some workloads have more to gain from a move to thewith the other workloads, though not as pronounced, have beencloud. This is often because they have a greater affinity for thevery promising so far. As of this writing, they are still evolving attributes inherent in the cloud model. They align with the stan-and will continue to do so as their respective implementations dardization, virtualization, automation and level of managementmature. 4. 4 Success in the cloud: Why workload mattersand hardware support that a cloud service provides. Theseworkloads can operate easily in a virtualized, automated cloudenvironment, where the infrastructure makes it possible toHigh Collaborationdynamically request services from a virtualized pool of hardwareand then automatically provision the required software stack and Development/ testresource capacity. Analytics Self-serve applicationWorkloads with the greatest affinity for cloud may be those development Storagethat are an excellent fit in terms of their potential gain and ease environmentof deployment. These workloads require little to no customiza-tion because they can work with and benefit from the clouds Gain Complex/critical Simple/non-criticalproductionproductionWebMailstandard catalog services. Figure 1 illustrates how several (hosted)common IT workloads fare on these two measures, based onIBMs research and experience with our own cloud implementa-tions. In general, the workloads that appear in the upper right ComplexERPComplex/criticalquadrant have proven to be the best fit for cloud computing. production (virtualized)LowWorkloads must be carefully analyzed to weigh potential gainsagainst how easily they can be deployed in the cloud, whetherpublic, private or hybrid. The fact is not every workload is the Deployment easesame in terms of its importance and cost to the organization, andLowHighthis can affect its outcome in the cloud. Some of the most criticalworkloads are so costly to the organization financially and opera-Figure 1. Workload affinity for cloud computing. Potential for gain andtionally that a move to the cloud has the potential to providedeployment ease are two key factors in the determination of a workloadsaffinity for and prospective success in the cloud.considerable benefit. Other workloads may be so highly opti-mized already that there is little to be gained from such a move.Similarly, legacy and heterogeneous applications, and workloadsWorkloads that are unusually complex may offer big potentialthat involve a high amount of data transfer, may be difficult tofor improvement, but they may require a high degree of custom-deploy in the cloud, relative to their anticipated gain.ization or application redesign to conform to the clouds archi-tecture. These kinds of workloads can prove too difficult,By contrast, self-contained applications may be easy to deployrisky or costly to move to the cloud. Certainly, the risk/rewardbut only offer moderate gains. Other workloads may be viewedprofile of any workload should be clearly understood before itas a good fit for cloud deployment simply because they pose lit-is deployed in the cloud. Consider complex ERP (enterprisetle risk to the organization from a security or other standpoint.resource planning) applications. The mission-critical nature of This is the case with collaboration workloads, for which there isthese production w