Solitaire Interglobal - Zift Solutions

---

S olitaire Interglobal Why Your Virtualization Matters How real companies maximize the value of their IT platforms “Virtualization has become one of the key components of our efficient IT department. When we switched to a virtualized environment, the projection was that it would save us somewhere between 5-8% of our overall IT budget. In the last three years, during which we have deployed virtualized environments in both production and non-production, the savings has been MUCH higher – closer to 38%. Without virtualization, our overall spend would be much higher, and our ability to support our expanding business much lower. It is the number one strategy to control costs and mitigate risk for us.” Financial Services CIO

A compelling shift in how IT supports the ever-escalating pace of business is taking place. The advent of the cloud has triggered this paradigm shift, irrespective of whether an organization actually deploys cloud technology or not. This fundamental change in the way that organizations do business has a corresponding evolution in the management of IT systems and resources. The operational structure is reforming, as the vision of the underlying relationship among computing, storage, IT staffing and business alters. Virtualization is the most significant force behind the paradigm shift, since it is built on the ability to direct and allocate resources, while separating disparate workloads. This capability to separate applications from one another, so resource prioritization can be effectively supported, is a prerequisite for cloud and the new vision. Cloud technology escalates the paradigm shift to a more intense level. The management tools incorporated into IT operations have also started to change significantly as tasks that were performed by multiple people, focused on discrete areas, are now executed by a single person. A toolset’s ability to increase pattern recognition and optimized presentation, while minimizing context switching, is key to efficient and reliable operations in the changed environment. One question for IT is “What is Virtualization?” This question may seem simple, but the answers are not the same across different virtualization methodologies. The clinical definition of virtualization is the creation of a simulated or “virtual” environment that is not tied to a physical or actual version of something, like an operating system, server, network connection, storage, information repository, etc. The ability of the virtualization method to create functional separation varies in both the resources that are virtualized and the clarity of separation that exists among all of the virtualized environments. Operating system (OS) virtualization has been around for a long time, starting as a mainframe mechanism decades ago. The virtualization of the OS has allowed a single platform to run multiple operating system images at the same time and also allows the system administrators to avoid wasting expensive processing power.

Telephone 847.931.9100 Website www.sil-usa.com

180 South Western Avenue, # 275 Carpentersville, Illinois 60110

Virtualization – Built on IBM POWER8 Technology

This goal has been maintained with the successive types and forms of virtualization manager (VM) software. Three areas of IT where virtualization is focused in the current market are resources for network, storage and servers. Each of these provides different opportunities and challenges for virtualization strategic selection. Network virtualization is a method of pooling the available network resources and then splitting those resources into subdivisions or channels. Each channel is independent from the others and can be assigned (or reassigned) to a particular server or device. Some virtualization methods allow this in real-time, while others require system restarts for changes in configuration. This management layer hides some of the complexities of arranging the resource allocation, but the difference in virtualization methods makes it important to understand the actual functions of management and allocation supported. This type of network virtualization is distinct from the virtualization of network bandwidth inside a server or storage device. Storage virtualization likewise involves the pooling of the physical resources across multiple devices into a simulated single entity. This is then managed by the virtualization architecture. The different virtualization methods support different functionality of network access into the virtualized storage, with some creating channels in the network and internal I/O, while others do not separate the resources. Server virtualization is the primary focus when IT looks at virtualization in the current marketplace. This arm of the virtualization effort also masks the resources available to the IT user, covering up most of the detailed information on where the actual work is being done. The primary goal of server virtualization is a drive to a more efficient and effective use of the organization’s IT computing resources, without the user needing to be constantly aware of complex location details. This goal relates directly to the cost side of both managing disparate workloads and maintaining workload separation. The secondary goal of leveraging opportunity is also considered in server virtualization, because virtualization strongly affects IT agility and the subsequent support of the core organization business. The three areas of virtualization (network, storage and server) are really reflected inside the actual architecture of the server. Server virtualization can include apportionment, separation and even isolation of computing capacity, memory, network and storage access, etc., all within the confines of the physical platform itself. Not all virtualization methodologies provide identical levels of resource management and security, nor does the control of all server resources exist in all vendor offerings. Many of the VM architectures merely separate the resources, rather than isolate them. This shortfall in the hypervisor component of the VM can expose additional security risk, as well as resulting in a less efficiently operating server, due to the “noisy neighbor” syndrome. Another question is why does an organization deploy virtualization? The short answer is revenue. The primary targets of virtualization are cost control and increased revenue. Each of the other goals that organizations state for virtualization usage can be tied, directly or indirectly, to this dual objective. If this goal is kept in mind, even things like stronger security, faster time to market, etc. that virtualization typically influences can be correlated with revenue targets and opportunities. © 2014, Solitaire Interglobal Ltd.

Page 2

Virtualization – Built on IBM POWER8 Technology

Virtualization Adoption and Use The deployment, adoption and use of virtualization are at the core of this study. The impetus for virtualization deployment tends to be a combination of competitive and cost-control concerns. From the business side, a CEO’s awareness of slipping market share, or declining profitability, coupled with the marketplace news of the impact of virtualization at other organizations creates a strong push for implementation. This is especially true for organizations where margins are slipping, personnel have been cut to dangerous levels and cost control is critical. The advent of virtualization within an organization radically changes the balance of overall task distribution among personnel, as well as the form and controls for capacity planning. As departmental management, i.e., marketing, logistics, etc., are striving to deploy infrastructures that avoid straining reduced operational staff loads, the use of virtualized environments helps to avoid incremental operations full-time staff equivalent (FTE). “We have had so many new roll-outs and deployments in the last two years that my staff has been suffering from massive burnout. We have had 4 times the normal turnover in staff, more sick time and more exhaustion errors than I can remember in 40 years in this business. About 9 months ago, we decided to start to really focus on the use of virtualization. Some of the driving force behind that was to prepare for possible cloud use, but mostly it was a strategy to try to get a better handle on our overwhelming workload. Overall, the strategy seems to be working. Some of the departmental lines are getting fuzzy, but the total staff load is getting better, even as we continue to have to respond to requests for faster and faster implementations. We have about a third of our applications in VMs, with another third planned for this next year and I fully expect the trend of efficiency to continue.” CIO large financial services organization

Virtualization efficiencies extend beyond the staff task time into the overall capacity requirements for the server platforms. This successful reduction in the size of the infrastructure can be viewed as an increase in efficiency. The form of the virtualization architecture plays a big part in the level of efficiency, as some vendors provide more thorough resource management than others. The accompanying chart shows the average efficiency increase reported by customers in this study, broken down by general categories of VM functionality. Where significant function combination options are not currently supported in the market, they do not appear on the chart, e.g., isolated CPU only. The summary of the reported efficiencies show that increased levels of control and separation produce higher realized savings in platform resources. This in turn is reflected in expenses, both capital and operational.

© 2014, Solitaire Interglobal Ltd.

Page 3

Virtualization – Built on IBM POWER8 Technology

The efficiencies from successful virtualization deployment do not end with the IT infrastructure. The way that IT staff is organized, responsibilities assigned and the overall load of task time is also undergoing a major shift. When personnel focus moves from the control and support of separate physical platforms to virtual ones, the lines between tuning of computing resources, storage and network start to blur. If IT operational activities have to be passed among different people, the lag time for that handoff introduces both a longer overall latency to the activity and more opportunities for inefficiency. A comparison of application deployment timeframes that include virtualization against those that do not was performed in a prior SIL, completed in April 2014 (Project ID: 9043.1008) and included results from over 110K organizations. The results support the general findings that the more comprehensive the virtualization methodology, the larger the effect on the organization. It also clarifies the relative value of the higher level of control in terms of agility. Another significant finding from this study was the correlated IT operational time that was required to deploy the reported application roll-outs. As would be expected, the overall average staff time was reduced, but the pattern of reduction is enlightening. The graphic visualization of the task time shows that the impact of virtualization functionality and architecture is significant. This part of the analysis shows the importance of the tooling that accompanies the virtualization method. An un-virtualized environment actually can be deployed with less overall staff time than some of the less complete virtualization architectures, primarily due to the tools that are available in those methodologies. This staff time requirement level can be tied directly to the functionality flow, with the less-integrated VM toolsets requiring extensive contextswitching from the operational staff. This negates many of the advantages that are possible with virtualization. When a single staff person has to either change their focus and cognitive context to use a separate tool to © 2014, Solitaire Interglobal Ltd.

Page 4

Virtualization – Built on IBM POWER8 Technology

accomplish the sequential tasks needed, the loss in time due to cognitive continuity and the increased possibility of error is significant. While the overall timeframe to implement is shortened for virtualized environments, this analysis shows that the staff time demand can actually be higher for those virtualization architectures that do not include robust tools. The tools’ functionality to increase pattern recognition and optimize presentation, while minimizing context switching is key to efficient and reliable operations in this changed environment. “This last year has been one that has seen a huge reorganization of our IT departments. Several technologies have caused ripple effects in the way that our IT management needs to group the personnel and actually what people we hire. The time and domain of the specialist seems to be fading, with dedicated area staff configurations, such as storage management, being reduced as our operational support staff starts to cover multiple disciplines. This would not be possible without the really excellent tools that we have. But thanks to those tools, the cloud, virtualization and some other general process changes, our overall staff requirements have been reduced by 21%.” Retail VP - Operations

Areas of Focus and Business Impact An increasingly critical requirement for any business operating today is rapid response to business need. With growing online customer interaction, agility has become the watchword for modern business. Lack of speed and throughput, as well as any shortfall in availability, are highlighted by missed opportunity and adverse revenue impact. A platform that supports rapid deployment, high availability, speed and consistent throughput has a measurable, and highly visible, benefit to the organization and has to be considered in the evaluation of any IT architectural component, including virtualization. Coupled closely with the focus on agility is security. The increasing availability of realtime and online interactions with customers also opens up larger vulnerabilities for businesses and customers alike. These areas of concern have to be addressed by the virtualization mechanism to be effective. While application security can be layered onto the individual application, the vulnerability of data, process and other intellectual components are not as easily protected and require other strategies. Even the application security is an incomplete and inconsistent protection, since applying interdiction controls on a case-by-case basis fails if any one of possible vectors are incompletely protected. Virtualization architectures vary considerably in their ability to both assist in the secured environment and in their ability to establish a good base for application, data and process protection. The primary differentiation lies in the degree of separation and isolation that is designed into the VM foundation. SIL draws a distinction between virtualization separation and isolation, although it could be argued that isolation is just a complete separation. However, the base architecture for a virtualization that provides real isolation is significantly different from one that ratchets up the degree of separation. Some virtualization methodologies have created strong hypervisor processes to keep the allocation routines, priority handling and other mechanisms separated, and the subject virtualized environments apart, but there is never a total uncoupling of resource and access pools. As the virtualization density of the separation-based architectures © 2014, Solitaire Interglobal Ltd.

Page 5

Virtualization – Built on IBM POWER8 Technology

increases, the overhead of the virtualization itself grows rapidly. The contrasting isolation architecture is designed to allow the hypervisor to oversee and move resources from one isolated environment to another. The environments are isolated to start with and the hypervisor connects them only to move resources to and from the isolated environments. The isolation architecture provides many advantages in both performance efficiency and security. While the efficiency comparison was documented previously, the security comparisons are just as significant. In the SIL Security Watch quarterly report, issued to SIL SW members for the second quarter 2014, the security incursions for the 2014 calendar year for January through June shows the wide disparity of the effects of the different classes of virtualization. The significant reduction in security incursions in the isolation architectures is a strong argument for considering this form of virtualization. The combination of the cost and performance effectiveness and efficiency of the VM, speed to market and the security aspects for the triumvirate of decision vectors have been shown to provide the basis for successful virtualization deployment. Virtualization Platform Considerations The selection of the optimal platform for deployment that includes virtualization must consider both technical and business factors, but the overall decision is a business one. Since the impact of platform selection is a difficult one to quantify on this level, IBM engaged Solitaire Interglobal Ltd. (SIL) to conduct surveys, gather data and perform analysis to provide a clear understanding of the benefits and relative costs that can be seen when organizations implement IBM Power Systems built on POWER8 technology as part of their virtualization IT architecture, incorporating IBM’s PowerVM as the virtualization method of choice. This analysis has been primarily directed at the value of platform use from a business perspective, so that business leadership can understand the benefit of the IBM POWER8 technology and PowerVM in organizational IT deployment and evolution. During this study, the main behavioral characteristics of software and hardware were examined closely, within a large number of actual customer sites (61,250+). These customers include organizations that have deployed virtualization architectures as part of their production environments. This group has organizations that maintain deployments that have been customized to support additional functionality and business process, as well as those that have been integrated with third-party or custom components. The information from these customer reports and the accompanying mass of real-world details is invaluable, since it provides a realistic, rather than theoretical,

© 2014, Solitaire Interglobal Ltd.

Page 6

Virtualization – Built on IBM POWER8 Technology

understanding of how the choice of platform can affect the organization’s costs, risk and strategic positioning in the current marketplace. In the collection and analysis of this data, a series of characteristics were derived. These characteristics affect the overt capacity, efficiency and reliability of the environment and its influence on operational and business performance. These characteristics have been examined within a business framework, since that is the perspective that is needed to make an informed decision. The business perspective encompasses a myriad of factors, including costs, staffing levels and other effects. These are the basis for selection of an optimal platform for an organization’s virtualization choices. Business Perspective Ultimately, IT and technology are designed to support business functions. So one of the primary perspectives of the study was the view of the technology by an organization’s business management, both executive and line-of-business. For the business portion of the analysis, the patterns of operations from the study organizations have been grouped into similar categories and then compared to identify their affect on business metrics. These metrics are: • • • • •

Customer satisfaction Total cost of ownership Staffing IT stability and reliability Agility (time-to-market)

Each of these business metrics has measurable and significant differentiation when the projected PowerVM on IBM POWER8 deployment solution is viewed. The more granular business metrics are those measurements that show how a specific measure of success is different in the general population of the implementers versus those that have deployed PowerVM. For further clarification, those situations where, Hyper-V, KVM, OVM, OVM for SPARC, VMware or Xen was the virtualization mechanism of choice have also been broken out. These metrics are fairly broad in coverage and touch on areas of financial consideration, as well as organizational quality. The metrics are presented with short definitions and the focused net effect of the VM deployment. In order to be meaningful across a variety of industries, all of the metrics have been normalized on a work-unit basis1, and categorized by levels of organization size (small, medium, large and very large). The base measure has been set by the medium company average, so that all other metrics are based on a variance from that standard set point. The implementations included in this study have been restricted to those implementations in production. Customer Satisfaction – Executive Management The ultimate metric on a successful implementation is customer satisfaction. SIL tracks this metric split out between the executive management and the operational input from

1 Work-unit basis has been defined using the published International Function Point User Group standards and are based on function point (FP) analysis.

© 2014, Solitaire Interglobal Ltd.

Page 7

Virtualization – Built on IBM POWER8 Technology

a customer, since the perspective of the customer may radically differ between those two groups. The satisfaction of the customer executive management with their IT systems tends to focus on the application, rather than the virtualization, although no application can work as well with a poorly configured or fragile virtualization method. That being said, the satisfaction with IT implementation and operation provides the most general metric for evaluation. This satisfaction rating was obtained from a large group of customers and provides a singular perspective on the overall success of virtualization deployment. While this is a subjective rating provided by high-level organizational management, it does provide the business’ actual perception of success.

The PowerVM-managed applications reflect in executive management satisfaction that is as much as 2.2 times better than the competitive options. The advantages seen by the reporting clients show increasing satisfaction in the applications run under PowerVM, much of which can be attributed to the number of complaints that the executives reported from their customers and users of those systems. The following chart shows the reported average monthly complaint count for the different platform groups. These complaints have been restricted to continued operational issues, and exclude complaints associated with missing and desired application functionality. “You [SIL] asked us for our satisfaction with our applications, servers and the virtualization that they are using. We have four different types of virtualization right now. Two of them are rock solid and never show up in our fire call reviews – the mainframe one and the one running on the new POWER8 boxes [PowerVM]. The applications on those boxes seem to only show up on my issue summary when there is a fight among the business groups and marketing over schedules. Other than that, they run, they don’t break, they don’t get hacked. Since that is not true of the other two, we will be phasing those out, even if their supposed cost was less. The security breach that messed us up last month was enough to pay for any amount of difference for years to come. I definitely do not want that repeated, every [sic] again.” Medium-sized manufacturing COO

© 2014, Solitaire Interglobal Ltd.

Page 8

Virtualization – Built on IBM POWER8 Technology

The PowerVM deployments show a relative reduction in complaints by as much as 88.17% when compared to the competitive offerings. While the specific customer complaints can be affected by management techniques, application design and other factors, the relative comparison is a legitimate indicator of how well the platform supports the processing at the organization. The three top reasons cited by reporting customers for the satisfaction were: • • •

Smooth running operation with little downtime and complaints Speed of implementation on IBM systems Consistency of service

The stability and resiliency of PowerVM deployments on POWER8 platforms are reflected in the low level of user complaints. Customer Satisfaction - Operational The operational perception of the customer, based on a variety of component metrics (e.g. support levels, communication, price, etc.), demonstrates satisfaction and success at the most generic level. This satisfaction metric is different from the overall satisfaction metric described earlier, in that the previous metric was gathered from the executive management level, while this metric examines the feedback from the operational side of the organization. This specific metric comes from information reported both by the IT departments and the line-of-business (LOB) groups.

© 2014, Solitaire Interglobal Ltd.

Page 9

Virtualization – Built on IBM POWER8 Technology

The satisfaction of the IT operational staff and the LOB with the PowerVM deployments reflect the reliability and resiliency of the platform as a deployment choice. The difference in operational management satisfaction favors the IBM solution by as much as 151.9% compared to the competition. The most highly cited reasons for the satisfaction were: • • •

Smooth running operation with little downtime and complaints Automated tools for management User interface efficiency

More than 98% of the reporting customers cited one or more of these three reasons for their satisfaction. Those organizations citing the user interface efficiency as a significant factor in their satisfaction differentiated between the simplistic tools that address individual tasks and the more workflow-friendly structure of the PowerVM tooling. Those that strongly favored the PowerVM tool structure consistently mentioned the complexity managing characteristics of the tooling (91.3%), and referenced the time savings and error reduction that the management software provided. Overall Expense (TCO) The cost perspective looks at the total cost to the corporation during a specific time period. This is normalized on three bases: employee, sales revenue and legal entity count, and contains expenses associated with up to a 3-month deployment preparation phase. These expenses span all of those included in the operational cost metric and are supplemented by expense contributions for physical plant, corporate overhead, longterm investments, etc. The TCO financial metric is more comprehensive than a straight operational metric. This metric should not be viewed in isolation, since extraordinary expense patterns for individual organizations may cause minor variance in the exact comparison values. For this reason, the comparison metric should be viewed as © 2014, Solitaire Interglobal Ltd.

Page 10

Virtualization – Built on IBM POWER8 Technology

indicative and providing a general range rather that an exact value. However, with the large number of contributing organizations, the data is sufficiently large that, combined with the other business metrics, this comparison helps to set an appropriate perspective.

The PowerVM-hosted applications shows lower overall expenses by as much as 81.16%, over a wide range of organization sizes. It should be noted that this TCO comparison should be viewed in conjunction with availability and downtime metrics. Since no cost has been associated with unavailability, each organization should factor in its associated downtime cost to the TCO metric provided here. The downtime metric can be found later in this document. For most customers, the cost of acquisition is higher with the POWER8 platforms than for the smaller Intel platforms, although sometimes only marginally. This disparity in cost levels is obviated when the level of virtualization and capacity demand increase. This switch in the defining metric from TCA to TCO happens in all situations eventually, but is more rapid in the larger deployments. Since the TCO holds true as a metric, well past the usefulness of the TCA, the TCO is used as the defining cost metric. The differential among the solutions is based largely on the lower expenses for the efficient deployment and the lower overall cost of the solution, including staffing. This is affected strongly by the scope of the virtualization deployment, with increased expenditure efficiency present as the complexity and size of the virtualization deployment increases. Customers of all degrees of deployment reported a consistent pattern of differentiation in three main areas: • • •

Lower staffing costs overall (due to tools, stability, etc.) More highly-leveraged platforms Lower datacenter costs (environmental, facility, etc.)

An interesting metric can be seen if the efficiency of scale (EOS) is examined for the virtualization options in this area. This measurement looks at the change in the normalized cost as the implementation increases in size and complexity in either the © 2014, Solitaire Interglobal Ltd.

Page 11

Virtualization – Built on IBM POWER8 Technology

physical deployment or the number of VMs. It reflects any efficiency that tools and management flow provide in a specific virtualization mechanism. Using the organization size as the driving principal, the TCO EOS trend can be summarized as shown in the graph below:

This graph shows the effects of the aggregation of the system efficiency. Detailed analysis shows that this correlates closely with the effective tooling that is built into the virtualization methodology. It counters an increase in staffing, with its attendant costs and complexity. While the IBM POWER8 architecture has incorporated a philosophy of granular and rule-driven scripting to reduce staff involvement and to extend the scaling efficiency, other VMs have concentrated simply on presentation layer functionality, which, while being both visually and intellectually appealing, does not result in the real results desired. The POWER8 implementation of PowerVM shows a clear implementation of the efficiencies of scale, which is counter to the competitive offerings. The top three sources of this advantage were reported as: • • •

Scripting capabilities of the software Efficient resource sharing Advantageous licensing structure

These factors have produced a realized savings in a blended cost per VM of 16.8% for PowerVM. Staffing An underlying factor that shows itself in many other areas is the effectiveness of the interface between the technical user and the infrastructure, including software, © 2014, Solitaire Interglobal Ltd.

Page 12

Virtualization – Built on IBM POWER8 Technology

hardware and operating system components, and the subsequent effect on staffing. The efficiency of any of the specific components that provide that influence on the user experience are difficult to break down into metrics other than in overly-detailed comparisons that lose their effectiveness by virtue of the degree of detail. Therefore, a general view of the full-time staff position equivalents was reviewed to provide a general metric for the platform comparison. These levels are those required to maintain a “gold standard” environment for each operating system group. Once again, in order to provide a level comparison field, the workload on the systems was normalized to identical levels. The set point for comparison was selected as the staff level for a medium-sized organization using VMware.

Since different virtualization methodologies have varying sets of implementation standards, it is important to keep the rigor of those standards in mind when reviewing the staffing. The noticeably lower staffing level for PowerVM deployment and use is directly attributable to an efficient unified workflow, as well as a substantially different mechanism to handle the allocation of virtualized resources. This is of special note as the organization increases in size or if an organization is on the path to a cloud service delivery model. The normalized staffing levels for PowerVM are smaller than those for the competitive offerings by as much as 66.0%. Research Note: Differences in the virtualization resource consumption were seen within all of the methodologies when the virtualization architecture was split out by the integrated storage platform and compared to the group average. Most significant of these were the virtualization methodologies of VMware, HyperV and PowerVM. The storage-driven variance for VMware was -9.2% to +4.6%, -14.3% to +5.6% for HyperV and -1.3% to +24.5% for PowerVM. The large variance for POWER8 and PowerVM resulted from efficiencies in the integration of platform and tools with the storage subsystems. The most efficient were those that were also IBM brand, but substantial efficiencies with other vendor technologies were reported for the POWER8 implementations.

© 2014, Solitaire Interglobal Ltd.

Page 13

Virtualization – Built on IBM POWER8 Technology

Staffing was further differentiated by the degree of cloud implementation within an organization. As cloud technology was further deployed, organizations reported higher efficiencies for the IBM POWER8 solution, with competitive advantages reaching as much as 4.5 times over other VM offerings. Another way to examine the staffing requirements is to look at which areas of activity consume the staff hours. For the purposes of this analysis a subset of the reporting organizations allowed SIL full time-motion data. This data was then analyzed to build a list of the top activities that the staff supporting virtualization performed. This occurrence analysis uses the frequency of the action to determine the weighting. Task Frequency Summary Rank 1 2 3 4 5 6 7 8

Description Check resources levels Reallocate and prioritize resources Setup VM environments Tune performance parameters Move applications - across environment types Setup new server and VM Move applications - within environment type Install patches and fixes

From a time perspective, the task list order changes, since some of the frequently performed tasks are simple and quick, while some of the other tasks take considerably longer. The ranking in this table is in order with the task consuming the most time first, the second next, and so on. Task Duration Summary Rank 1 2 3 4 5 6 7 8

Description Move applications across environments Tune performance Move applications within environments Setup VM Reallocate and prioritize Setup new server and VM Install patches and fixes Check resources levels

If the top three task areas are examined from a relative time consumption perspective among the virtualization options, an interesting pattern appears. This summary compares a normalized environment against the set point, which in this case is set by VMware at a medium-sized organization.

© 2014, Solitaire Interglobal Ltd.

Page 14

Virtualization – Built on IBM POWER8 Technology

There is a radical difference in the amount of time spent on the top three staff time usage tasks when PowerVM is included in the analysis. The PowerVM advantage is as much as 83.64% in these most heavily performed staff tasks. Part of this difference can be correlated to the workflow design within the PowerVM management tools. The overall context switching was significantly reduced in frequency, by as much 87.2%, when compared to the competitive average. This translates into an IT support person performing virtualization tasks having to switch workflow direction or open additional screens 5/6ths less frequently. The lower number of context switches results in fewer mistakes and faster task completion, while the substantial scripting advantage of the PowerVM tooling further reduces the overall task time. Both the reduced context switching and the script capabilities of the PowerVM toolset are also visible in the reported learning curve timeframes. The data from the reporting organizations included the interval of time that a staff member needed to be fully functional in each of the virtualization products. This was not a timeframe for expertise, which has many different metrics, but a base one of adequate performance, obviating the need for training supervision. The reporting organizations provided learning curve data that focused on an IT support person being able to handle a full lifecycle of virtualization, from creation through non-production environments, all the way to full production deployment and resource management. This is significantly different than being able to execute simple tasks on a simple tool. The metric encompasses competency, knowledge and the ability to operate independently. This information is shown in graphic form in the following chart.

© 2014, Solitaire Interglobal Ltd.

Page 15

Virtualization – Built on IBM POWER8 Technology

The learning curve on PowerVM takes a significantly shorter time to competence, with the most frequently stated reasons of: • • •

Robust management tools Optimized workflow Quality of education and training

The faster ramp-up time of the PowerVM virtualization method is as much as 2.39 times faster than the others in this study. This faster time to competence can be critical for organizations, as they deploy new virtualization efforts. Efficiency in the staffing area, including the ramp-up time is especially important in the growing market of managed service providers (MSP). Since the basic profitability of these MSPs offering either direct or cloud services for a myriad of customers hinges on efficiency of staff execution, any savings in the staff that supports a growing array of VMs is important. When the MSPs that were included in the study are isolated and examined, the differentiation of the POWER8 and PowerVM architectures is extremely evident. Those MSPs that are utilizing this combination report less than 28.4% of the staffing hours per customer VM and less than 25.4% of the cost per VM as other architectures. All of the MSPs that utilize POWER8 and PowerVM have indicated that they are planning on expanding their footprint of this architecture. “Our VMs are the basis of our business operations. The fact that POWER8 with the PowerVM virtualization is more than three times more cost efficient for us is the reason that we are expanding into that platform as fast as we can. While the x86 boxes are cheaper to buy, they are not cheaper to operate and maintain. We have had some learning curve on this, but are happily on our way to much more profitable operations with the IBM PowerVM components.” Large MSP CIO

© 2014, Solitaire Interglobal Ltd.

Page 16

Virtualization – Built on IBM POWER8 Technology

IT Stability, Risk and Reliability Risk is composed of many factors. It includes the stability and reliability of the platform, as well as the chances of platform failure. IT stability and reliability metrics include all downtime, both planned and unplanned. The dependability of the implementation is a combination of the individual reliability of each component, along with the quality and effectiveness of the actual implementation. As such, both the planned and unplanned outages affect the overall usability of the total system. SIL views availability as a combination of all outages, i.e., network, hardware, OS, DBMS, etc. The number of outages has been normalized for a 10-platform operation, with both planned and unplanned outages included. Where virtualization has been included in the architecture, each of the virtualized environments has been considered as a separate platform. Each of these outages takes valuable access time away from the corporate resources. The following chart shows the percentage of time that those outages represent and includes all forms of unavailability, irrespective of source.

As shown above, there is a substantial indication of how the PowerVM virtualization contributes to both stability and reliability of an organization’s implementation, due to the combination of high performance and native resilience. The three most cited sources of the high availability from customers are: • Limited need to reboot the full platform • Fewer system patches and updates required • Responsiveness of technical support It should be noted that the practices of the individual organization when viewed from a best practices perspective makes a difference in the amount of planned downtime. However, the overall trend in availability is a definite indicator of platform stability. The cost of that availability is difficult to articulate, primarily because such a cost estimate has significant subjective components. However, a quick analysis of the © 2014, Solitaire Interglobal Ltd.

Page 17

Virtualization – Built on IBM POWER8 Technology

customer-reported financial impact of outages yields a general metric that provides some interesting insights. SIL considers risk to be comprised of three components: • • •

Percentage chance of component failure Percentage chance of budget or timeframe overrun Potential exposure, expressed as a percentage amount of overall budget or timeframe overrun

These three percentages are added to form the overall risk factor for a scenario. The risk factor summary for the platform scenarios is shown below.

This graph shows that there is demonstrated risk mitigation from the general operations experience when using PowerVM. The risk exposure for PowerVM is significantly smaller than the competition, with PowerVM deployments showing significantly less risk that has been reported for other virtualization methods. These competitive VM architectures demonstrate higher risk by as much as 13 times that of PowerVM. Much of this lower risk can be attributed to the high resiliency of the deployment and increased efficiency of the resource allocation within the virtualization component itself, which significantly lowers the risk of component failure. A more granular chart shows the detail of the different risk contributions. It should be remembered that these are averages across all of the reporting organizations that use the particular VM architecture.

© 2014, Solitaire Interglobal Ltd.

Page 18

Virtualization – Built on IBM POWER8 Technology

The PowerVM deployments have such a small risk profile that the risk contributions are too small to delineate the separate contributions in this graphic form. Agility Agility is defined as the average number of calendar days from the start of an initiative to the start of full production operations for a project. This is NOT staff days or hours, but the actual calendar span, including all weekends, holidays, etc. All of the contributory factors, such as staffing and reliability, radically affect the speed in which a company can move a business concept from inception to market. This nimbleness is a key element of increasing market share and continued corporate viability. While the performance metrics were gathered on the production systems, additional measurements were also collected to track the amount of time that the systems took to move from initial conception to full production implementation. The results demonstrate a significant increase in agility when platforms running PowerVM virtualized environments were used. This increase in agility has been reported to be as much as 63% faster for the PowerVM systems when compared against the overall study group. This translates into a faster time-to-market for business initiatives. The comparison is intended to be evocative and not quantitative, since other critical success factors, such as management methodology, resource availability, etc., can enter into this picture.

© 2014, Solitaire Interglobal Ltd.

Page 19

Virtualization – Built on IBM POWER8 Technology

It is apparent from the reported data that there is a definite agility advantage to using PowerVM-deployed systems as compared to the overall experience, especially when organization standards for production system promotion are comprehensive. When asked for specific sources of the agility, the most frequently cited reasons from customers were: • • •

Robust tools set for management over multiple instances Ability to easily shift resources to accommodate new implementations Speed of movement from non-production to production environments

The differences in agility can be substantial, with the PowerVM showing faster deployment times by as much as 3 times compared to other VM methods. This means literally that a well-managed and optimally configured IBM PowerVM system is directly associated with faster time-to-market and more rapid response. Technical Perspective One of the main perspectives for this analysis is from the viewpoint of the IT professional. Since IT needs to understand the underlying architecture and important characteristics of any technology, this perspective tends to focus primarily on the objective understanding of what a PowerVM deployment can contribute and will require. This understanding encompasses some basic performance characteristics and operational challenges. System Efficiency The ability of the system to fully utilize its resources is a significant technical metric of the value of a particular component in the infrastructure. While one of the key points of virtualization is the amount of efficient usage that can be applied to the extant organizational platforms, an examination of the actual production deployment patterns © 2014, Solitaire Interglobal Ltd.

Page 20

Virtualization – Built on IBM POWER8 Technology

shows some interesting practices. The density of implementation is shown in the graph below.

The number of VMs per platform for several of the virtualization methodologies varies significantly from the general group. The PowerVM (POWER8 architecture) deployments show a substantially higher number of VMs per platform consistently across the organization sizes, with the gap widening as the size of the deployment increases. This higher level of efficiency shows as much as 12 times the VM density for PowerVM as the remainder of the analysis group. When the reasons for setting the VM density were analyzed, the top three were: • • •

Acceptable risk levels (71.3%) Platform performance constraints (70.2%) Application vendor constraints (18.5%)

The resource utilization per VM provides an interesting view into the technical considerations in this area. The ability of the virtualization method to move resources from one VM to another also comes into play. Effective sharing of resources intra-VM allows the virtualization method to achieve higher levels of overall utilization and load. In comparisons for this type of metric, the average system utilization is normalized based on the work executed inside of a VM and the cost of a normalized work unit is derived. The cost of this work is then normalized against the set point of a VMware medium-sized implementation.

© 2014, Solitaire Interglobal Ltd.

Page 21

Virtualization – Built on IBM POWER8 Technology

Each of the different virtualization methods has a characteristic footprint in relation to its memory and process architecture. While some of the methods show a trend of increasing memory consumption as the deployment scales up in size, the PowerVM deployments show an increasing efficiency that produces lower capacity consumption per VM as the deployment grows in size and complexity. The usage levels of each VM show clear efficiencies for PowerVM in resource utilization – an important consideration if fully leveraged platforms are desired. This advantage shows reduced resource consumption, with the POWER8 PowerVM using as little as 41.75% of resources needed by competitive offerings for the same workload. The cost per work unit for virtualized environments is an indicative metric for those planning a fairly complex environment. In this situation, the resiliency of the underlying architecture is also a substantial contributor to the efficiency of the virtualization methodology. The cost of deploying each VM is another metric that seems to span both business and technical. These costs include the average cost of platform resource and staff time, but exclude the actual application cost. This is especially important for organizations that have active and volatile non-production environments, since the change in those environments is far higher than that of normal production. When the deployment cost is examined, the comparison is extremely interesting, as can be seen in the chart below. Once again, these costs are normalized against the set point of a medium-sized VMware deployment.

© 2014, Solitaire Interglobal Ltd.

Page 22

Virtualization – Built on IBM POWER8 Technology

The cost per work unit for virtualized environments is an indicative metric for those planning a fairly complex deployment. In this area, the PowerVM advantage is as much as 95% cheaper than other options. “..For the first time in my recollection, we have had our budgets show an excess on a project. The virtualization project is showing a decreasing per environment [VM] cost as we develop the business rules and scripts. The PowerVM software is the main reason, I am told. The new POWER8 boxes have been deployed faster and cheaper than any other project in the last 6 years. And each wave seems to get a bit faster and cheaper.” Director of Operations for a large healthcare organization

Security Security is an important part of any virtualization solution, since virtualization concentrates security topology more densely. With the ability to create virtual machines within the same physical platform, the definition of IT security starts to evolve into more than simple access security. The concept of sidewise hacking, where access from one VM to another is broken, like blasting through the walls of an apartment to another within the same building, is a topic of discussion for security personnel everywhere, as hackers target virtualization architectures. The protections that the VM software provides have to cover a wider variety of access points than are necessary for security at a whole platform level. In this situation, control over all aspects of processing need to be in place. Many government and secure installations require isolation for their virtualized environments. This protection is required to span the allocation and handling of the main IT spheres: I/O, network access, memory management and overall normal execution access. This is where PowerVM has a significant advantage, since the other virtualization architectures are based on a separation foundation, rather than an isolation architecture. This intrinsic difference in virtualization approach poses a much more difficult barrier for hackers to breach. The imbedded encryption that runs the © 2014, Solitaire Interglobal Ltd.

Page 23

Virtualization – Built on IBM POWER8 Technology

virtualization interaction within PowerVM keeps the isolation of the different partitions complete, protecting not only the encapsulated application, but the other applications sharing the same physical platform. The fundamental difference has a large impact on virtualization within a single company, within a single cloud installation and also for those MSPs that offer cloud services. Increasing exposure to security incursions at cloud providers has been publicized many times over the last year. In these incidences, billions of dollars of customer-reported damage have been documented, with the continued fallout of the incursions echoing long past the initial attack. The most notable of this type of contagious theft is where hackers attack law firms handling an organization’s patents. When the patent application is stolen (often undetectably), the hackers then get the patent filed before the organization that has invested time and substantial money into the research and testing. This theft is only apparent after a time delay, and has exceeded $100 billion in loss. While any organization can have an incursion that relates to misuse of password protection, excluding that type of procedural control incursion, PowerVM has currently no reported incidences of a break in any of the VM security access points, making it unique in the analyzed group covered in this study. Cloud Integration The maturing cloud deployment landscape provides both opportunities and dangers to the implementing organization. The exact form of the cloud deployment affects the challenges, whether that deployment be for public, private, community or hybrid cloud. With organizations avoiding both the expense of the private cloud and the exposure of a public cloud deployment, an increasing number of hybrid and community clouds are being implemented. The hybrid deployments carry a variety of vulnerabilities by the merger of various levels of security, while the community clouds have increased exposure due to the common usage by a wide variety of different organizations. The integration with cloud services is a very common initiative in today’s marketplace. With that in mind, part of the study looked at organization deployment on the cloud and the link to virtualization strategy. In this area, there were several points of analysis. The first of these is the customer use of cloud, split out by virtualization method.

© 2014, Solitaire Interglobal Ltd.

Page 24

Virtualization – Built on IBM POWER8 Technology

The second perspective is which customers have selected their virtualization with cloud deployment (either current or planned) in mind. Of those responding, more than 89% said that cloud was a consideration in selecting their virtualization method. Both of these perspectives have created an interesting viewpoint into the cloud movement. Since cloud architecture is really a further form of virtualization, the selection of the architectural strategy says a lot about which technology has the robustness to position an organization for the future. In this area, PowerVM clearly has a strong presence. This presence is especially true for cloud service providers, where the combination of lower cost per VM, more effective security and better management tools make the POWER8 with PowerVM a prime candidate for consideration. Conclusion The Solitaire Interglobal Ltd. analysis of virtualization methodologies shows that there is a substantial advantage to incorporating the IBM POWER8 PowerVM offering within an organization’s IT architecture, based on a broad set of business and performance metrics. The advantages that accompany this inclusion increase the effectiveness of application deployment and translate to real-world positive results, experienced and reported by the businesses in this study. The study has identified critical business and performance metrics that can be used to understand the advantages and key strategies that will help an organization to choose the optimal deployment platform. While success can be measured in different ways and looked at from varying perspectives, it could be said that the bottom-line measurement of deployment success is overall customer satisfaction. Customer satisfaction incorporates a wide variety of qualitative and quantitative components, yet it is the simplest summary of how well a deployed system has met organizational expectations. © 2014, Solitaire Interglobal Ltd.

Page 25

Virtualization – Built on IBM POWER8 Technology

As outlined in the analysis, the customer satisfaction with the IBM POWER8 PowerVM choice is high, both from a business and a technical perspective. The economic benefits of the virtualization choice are also apparent in the control of overall expense. PowerVM is associated with significantly lower costs. The strong virtualization functions built into the PowerVM offering make a measurable difference. These functions provide the ability to sweep capacity resources to targeted processes, and result in the need for fewer overall system cycles. Coupled with the allocation automation, personnel time, hardware, software, and personnel costs can all be minimized. This produces efficient application deployment and cost-effective expenditures, while displaying a risk profile that is substantially lower than the other solutions examined in this study, due to the fundamental isolation functions of the PowerVM product. A few of the highlighted findings can be seen in the quick summary below. Quick Summary Category Customer Satisfaction Total Cost of Ownership (TCO)

Commentary The more complex or volatile the environment, the more all aspects of customers reported high satisfaction with PowerVM. POWER8 architecture with PowerVM demonstrates lower TCO by as much as 81.16% compared to competitive offerings.

Staffing

The normalized staffing levels for PowerVM are smaller than those for the competitive offerings by as much as 66.0%.

Risk

The reported risk of deployment is considerably better for PowerVM users, with the competition showing as much as 13 times the risk of POWER8 architecture and PowerVM.

Agility

PowerVM users are reporting faster deployment times by as much as a factor of 3.

System Efficiency

Resource usage consumes as little as 41.75% of the resources need for other virtualization, due to lean VM overhead and efficient operations. PowerVM supports all forms of control and isolation, including those required for highly secure implementations, separating resources for memory, network, I/O and access.

Security

Quick Byte PowerVM shows a strong support for changing customer needs. Better cost containment and operational predictability comes with PowerVM. Powerful scripting and workflow lets PowerVM leverage the efficiencies of scale. Flexible and powerful functions to share resources greatly lower the risk of deployment. A well-managed PowerVM system can be directly associated with faster timeto-market. Do more with less with PowerVM. No reported successful VM hacking in PowerVM.

These key findings are all substantial reasons to consider POWER8 platforms using Power VM for an organization’s IT infrastructure. The study metrics show an increase in the effectiveness of the IT deployment and translate to real-world positive results experienced and reported by the businesses in this study. Overall, critical effects on staffing, security, integration and satisfaction make IBM’s POWER8 PowerVM a strong contender for an organization’s virtualization, and platform, choice.

© 2014, Solitaire Interglobal Ltd.

Page 26

Virtualization – Built on IBM POWER8 Technology

Solitaire Interglobal Ltd. Solitaire Interglobal Ltd. (SIL) is an expert services provider that specializes in applied predictive performance modeling. Established in 1978, SIL leverages extensive AI technology and proprietary chaos mathematics to analyze prophetic or forensic scenarios. SIL analysis provides over 4,500 customers worldwide with ongoing risk profiling, performance root cause analysis, environmental impact, capacity management, market trending, defect analysis, application Fourdham efficiency analysis, organizational dynamic leverage identification, as well as cost and expense dissection. SIL also provides RFP certification for vendor responses to government organizations around the world and many commercial firms. A wide range of commercial and governmental hardware and software providers work with SIL to obtain certification for the performance capabilities and limitations of their offerings. SIL also works with these vendors to improve throughput and scalability for customer deployments and to provide risk profiles and other risk mitigation strategies. SIL has been involved deeply in the establishment of industrial standards and performance certification for the last several decades and has been conducting active information gathering for the Operational Characterization Master Study (OPMS) – chartered to develop better understanding of IT-centric organizational costs and behavioral characteristics. The OPMS has continued to build SIL’s heuristic database, currently exceeding 160 PB of information. The increased statistical base has continued to improve SIL accuracy and analytical turnaround to unmatched levels in the industry. Overall, SIL runs over 150,000 models annually in support of both ongoing subscription customers and ad hoc inquiries.

Methodology Notes In order to understand the impact of IBM POWER8 architecture and PowerVM platforms as a key part of an organization’s IT infrastructure, a significant number of deployments were examined. The relative degree of difference in operating behavior for each factor, i.e., total number of outages, etc., was then compared to understand the net effect of the respective combinations. The effects were observed in general performance and capacity consumption, as well as other business metrics. The approach taken by SIL uses a compilation and correlation of operational production behavior, using real systems and real business activities. For the purposes of this investigation, 61,250+ environments were observed, recorded and analyzed to substantiate the findings. Using a large mass of customer and industry experiential data, a more accurate understanding of real-world behavior can be achieved. The data from these systems was used to construct a meaningful perspective on current operational challenges and benefits. The reported behavior of the systems was analyzed to isolate characteristics of the architecture from both a raw performance and a net business effect perspective. All input was restricted to those organizations using systems in versions that were current in calendar years 2013-2014. Since many of the components in this environment have releases at staggered points in time, only those components that were either the current version or a -1 version based on those calendar restrictions were included in the study. Additional information on the methodology and study diversity can be found in additional methodology notes at the end of this document. In a situation such as that presented by this study, SIL uses a methodology that incorporates the acquisition of operational data, including system activity information at a very detailed level. It should be noted that customers, running on their production platforms, provided all of the information. It is essential to understand that none of the data was captured from artificial benchmarks or constructed tests, since the value in this study comes from the understanding of the actual operational process within an organization, rather than the current perception of what is being done. Therefore, these sites have tuning that is representative of real-life situations, rather than an artificial benchmark configuration. Since the focus of this analysis was not to tightly define the differences among different minor variations of operating system or hardware, the various releases were combined to show overall architectural differences. This provides a more general view of architectural strategy. In order to support the comprehensive nature of this analysis, information from diverse deployments, industries, geographies, and vendors were obtained. In any collection of this type, there is some overlap that occurs, such as when multiple vendors are present at an organization. In such cases, the total of the discrete percentages may exceed 100%. Those organizations with a multi-layered deployment, such as multiple geographical locations or industrial classifications, have been analyzed with discrete breakouts of © 2014, Solitaire Interglobal Ltd.

Page 27

Virtualization – Built on IBM POWER8 Technology

their feedback for all metrics. Additional filtering was performed to eliminate those implementations that substantially failed to meet best practices. Since the failure rates, poor performance and high costs that appear in a large number of those implementations have little to do with the actual hardware and software choices, these projects were removed from the analytical base of this study. The industry representation covers manufacturing (25.59%), distribution (12.49%), healthcare (14.74%), retail (5.43%), financial (20.50%), public sector (3.09%), communications (17.69%) and a miscellaneous group (0.47%). The geographies are also well represented with North America providing 39.63% of the reporting organizations, South and Central America 12.65%, Europe 24.57%, Pacific Rim and Asia 20.61% Africa 2.29%, and those organizations that do not fit into those geographic divisions reporting 0.25% of the information. Since strategies and benefits tend to vary by organization size, SIL further groups the organizations by the categories of small, medium, large and extra large. These categories combine the number of employees and the gross annual revenue of the organization. This staff count multiplied by gross revenue creates a metric for definition that is used throughout the analysis. In this definition, a small organization could be expected to have fewer than 100 employees and gross less than $20 million, or a value of 2,000, e.g., 100 (employees) X 20 (million dollars of gross revenue). An organization with 50 employees and gross revenue of $40 million would have the same size rating, and would be grouped in the analysis with the first company. The classifications used by SIL use thresholds of 2,ooo (small), 10,000 (medium), 100,000 (large) and 1,000,000 (extra large). The information in this study has been gathered as part of the ongoing data collection and system support in which SIL has been involved since 1978. Customer personnel executed all tests at SIL customer sites. The results of the tests were posted to SIL via the normal, secured data collection points that have been used by those customers since their SIL support relationship was initiated. As information was received at the secure data point, the standard SIL AI processing prepared the data in a standard format, removing all detailed customer references. This scrubbed data was then input to the analysis and findings.

Attributions and Disclaimers IBM, and POWER8 and PowerVM are trademarks or registered trademarks of International Business Machines Corporation in the United States of America and other countries. UNIX is a registered trademark in the United States of America and other countries, licensed exclusively through The Open Group. Other company, product and service names may be trademarks or service marks of others This document was developed with IBM funding. Although the document may utilize publicly available material from various vendors, including IBM, it does not necessarily reflect the positions of such vendors on the issues addressed in this document.

POL03210-USEN-00

© 2014, Solitaire Interglobal Ltd.

Page 28