The Business Case in Virtualization www. infotech. com Impact Research 1
Table of Contents 1. 2. 3. 4. 5. 6. 7. 8. 9. Executive Summary…………………………. . Detailed Fit Indicators………………………. . Benefit Drivers……………………. . . . Common Pitfalls…………………………. . Making the Virtualization Business Case……………. . Cost Avoidance…………………………… Cost Savings………………………. . . …. Necessary Expenditures………………………. Creating the Plan…………………………. www. infotech. com Impact Research 3 6 13 16 21 23 26 29 30 2
Executive Summary www. infotech. com Impact Research 3
Executive Summary Mid-market enterprises are in the sweet spot to achieve cost savings from virtualization u Despite being a very powerful technology, virtualization benefits are more readily realized in certain types of enterprises. u Two key factors are leading indicators of greatest success: Size of Enterprise Number of Servers (# of employees) (# of Supported Physical Servers) • Enterprise size affects the role of IT, the materiality of expenditure and savings, and comfort with new technologies. • Very large companies may experience more difficulty in implementation due to established bureaucracies. • Thirty servers or more leads to the greatest and most demonstrable cost savings and benefits when taking implementation costs into consideration • Very small companies have trouble justifying the investment. Sweet Spot: ~100 to ~5, 000 Sweet Spot: 15+ If there is organizational fit, virtualization does provide most vendor-stated benefits. www. infotech. com Impact Research 4
Executive Summary The virtualization business case will be rooted in hardware savings u Implementation of Virtualization does not affect the number of applications, but rather the number of physical boxes in use. u Thus, the majority of pure cost savings are linked to reduction in physical boxes. u Cost avoidance and incremental business benefits exist in areas of Disaster Recovery (DR)/Business Continuity (BC) and efficiency, but are not always as tangible. Business Case Value 1. Reduction in one-time and ongoing hardware acquisition costs 40%-75% 2. Reduction in hardware-related maintenance costs 25%-50% 3. Efficiency benefits from increased manageability Acquisition Cost Savings Monthly Recurring Savings 4. DR/BC enabled by removing need for homogeneous hardware www. infotech. com Impact Research 5
Evaluating Your Company’s Fit www. infotech. com Impact Research 6
Detailed Fit Indicators In the context of virtualization, success means achieving all three set benefits TANGIBLE BENEFITS: Achieved the stated business case financial objectives, including one-time and ongoing cost savings and on-time implementation completion. SUCCESS INTANGIBLE BENEFITS: STRATEGIC BENEFITS: Recognized value of virtualization and have plans in place to leverage technology for future growth. Recognized feature benefits inherent in virtualization and exhibited improved efficiency. www. infotech. com Impact Research 7
Detailed Fit Indicators It’s necessary to look both within and outside of IT to accurately assess fit u Virtualization can be implemented in any enterprise; however, success (as defined on the previous page) of that implementation depends on a number of factors, broken down into three categories: Company Characteristics IT Department Characteristics Server Infrastructure Despite being a “cool” technology, perform adequate due diligence in assessing fit. www. infotech. com Impact Research 8
Detailed Fit Indicators Company Characteristics u u Virtualization is really an infrastructure initiative that can bring some level of benefit to any size of company. Not Using Although adoption penetration of virtualization is greater in larger companies, the proportion of their environment that is virtualized is much less significant. Planned/POC Using The greatest benefit can be seen in small and mid-sized organizations. % of Servers Virtualized u Small and mid-sized companies make the most use of virtualization technology www. infotech. com Impact Research 9
Detailed Fit Indicators Company Characteristics u Smaller companies have seen greater benefits in the implementation of virtualization Being a more innovative IT solution, small to midsized companies have significantly better success in implementation for the following reasons in order of impact (shown as Harvey Ball on right): Company Size (# Employees) 1 -100 101 -250 1 Greater Executive Comfort with Virtualization • Greater Executive Comfort with Virtualization: Comfort tends to be significantly higher in smaller organizations, in a large part due to the less complex political climate and ease of buy-in into the potential savings from a virtualization implementation. Further, smaller organizations tend to be less risk averse to new technologies than larger organizations. 250 -500 500 -1000 1001 -5000 5001+ Virtualization works, but is less compelling. Sweet spot. Virtualization works, but is more difficult to implement and usually of reduced scope. 2 • Relaxed Formal Approval Process: No Need for a Approval Process Flexibility of a smaller company, in most cases, eliminates the need for a formal approval process, which can introduce risk by necessitating education of a broader audience on the workings of virtualization. 2 • Appetite for Incremental Cost Savings IT Department Autonomy: Incremental savings resulting from benefits, such as maintenance efficiencies and power savings, have proportionally more impact on a small company (who feels the impact immediately) than a large enterprise. 3 • ITReduced Political Complexity: Department Autonomy Across organizations interviewed, the IT departments of smaller companies had more latitude to make decisions regarding virtualization and related infrastructure topics. 3 www. infotech. com Reduced Political Complexity Although never absent, fewer established (and isolated silos) and a flatter organization cause more pragmatic decision-making. The IT strategic vision can be politically diluted in larger companies. Impact Research 10
Detailed Fit Indicators IT Department Characteristics IT’s commitment to virtualization is key to reaping maximum benefits u IT departments’ current infrastructure and commitment to virtualization play a key role in determining success of an implementation. u Three characteristics are indicative of success: 1 • IT Comfort with Virtualization: Much like executive comfort with the technology, the technical team must be willing to move away from physical servers. 2 3 • High Proportion of Serverbe Virtualized: Proportion of Servers to Virtualization Value in Virtualization is achieved when more, rather than less, of the server infrastructure is migrated. Migrating 50% or more of the environment is a good indicator of success. • Material Value of Power Savings: Facility space and power savings are two great benefits of implementing virtualization. If the IT organization views these as material, given their current environment and costs, then it indicates a good fit for the technology. www. infotech. com Impact Research 11
Detailed Fit Indicators Server Infrastructure u If you have 15 or more servers and plan to buy new virtual host machines, you’re in business! The physical servers that exist in an enterprise are an important component of determining the fit of virtualization within the environment. Please see Page 23 1 15 or More Physical Servers • 15 or more Physical Servers: for more details… The break-even point for virtualization of a small number of servers (considering hardware and software only) is roughly three servers. The value increases as the number of servers grow. Once all other costs are incorporated, 15 servers is the threshold to achieving greatest value. In short, the more, the better. 3 • Acquisition. Low-Utilization Applications: Number of of New Physical Servers Reuse of physical servers to act as host machines is a great way to increase the utilization of those machines; however, for most implementations, purchasing new hardware will extend the life of host machines and could provide additional scalability. 2 Number of Low-Utilization Application Servers • Acquisition of New Physical Servers: Quick wins can be achieved by virtualizing lowutilization and low-complexity server applications like DNS and Web-servers. Critical applications are certainly candidates as well, but require more time to test thoroughly and may extend the payback timeframe. www. infotech. com Impact Research 12
Benefit Drivers www. infotech. com Impact Research 13
Benefit Drivers There are real savings resulting from virtualization implementation u Virtualization implementations bring benefits that span both tangible and intangible benefit areas: Cost Avoidance Cost Savings • • • Cost Savings (spending less) Cost Avoidance (not spending more) Intangible Benefits (not quantifiable) Many clients interviewed realized ongoing savings of 30 -50% DECREASE HARDWARE ACQUISITION COSTS by buying and setting up fewer servers Please see Page 23 for more details… SHRINK YOUR FOOTPRINT by decreasing rack space and cutting back on cooling/electrical costs Please see Page 25 for more details… ENABLE AUTOMATIC FAILOVER to improve/enable DR and BC operations Please see Page 26 for more details… HELP BUSINESS GROW FASTER WHILE KEEPING STAFF CONSTANT by making staff more efficient and repurposing/reallocating freed resources www. infotech. com Impact Research Please see Page 27 for more details… 14
Benefit Drivers Intangible benefits can deliver as much value as hard cost savings u. Shortened Timelines for Development Projects Improved Manageability & Flexibility u Virtual servers can be monitored and managed from one console and allow for capabilities such as failover and snapshot image backups. Features like these improve the efficiency of system administrators while maintaining, or improving, stability. u. Enabling Virtualization of Desktops Intangible Benefits Improved Performance u Products like VMware allow for resource pool management, which dynamically load balances virtual servers across physical hosts. The result is more consistent performance and less manual intervention. Many companies are now virtualizing desktops to manage security (by allowing remote access to desktop images) and image integrity (through the ability to refresh a desktop image when necessary). Separate Hardware from Software when Upgrading u Controlling Server Sprawl u Management of virtual servers should be accompanied by a cultural change that reduces or eliminates the purchasing of physical servers outside of the IT department. The IT department will provision virtual instances as required and hold the power to make decisions over incremental purchases. www. infotech. com Development projects are often slowed by the availability of development and testing environments. With template-like functionality, virtual servers can be brought up and refreshed from a single console within minutes to meet project requests. An interesting benefit to organizations running large, demanding applications on servers that require frequent hardware upgrades is that no re-installation or reconfiguration is required upon hardware changes. u. Reduce the “Mess” Perhaps not as salient, reducing cabling by five- or eight-fold does significantly reduce the possibility of human error in unplugging the wrong cable. Impact Research 15
Common Pitfalls www. infotech. com Impact Research 16
Common Pitfalls Heed the warnings of your peers (continued) u Business case pitfalls u Be wary Be Wary of of including FTE headcount Reduction Including FTE Headcount reduction Companies we spoke to had found improved efficiency in their maintenance staff, but this is rarely sufficient to remove resources altogether since, in most cases, the system administrators perform other roles in the IT department. Quell Your Expectation of HUGE Consolidation Number of Virtual Machines Quell your expectation of huge consolidation. Consolidation ratios mentioned by the vendors differ from actual client experiences. 12 What vendors report… 9 6 What to expect… 3 0 Per Core www. infotech. com Per Processor Impact Research 17
Common Pitfalls Heed the warnings of your peers (continued) u Implementation warnings Executives and Staff Need toto have Faith u Executives and Staff Need Have Faith Don’t underestimate the culture shift required to remove physical servers and replace them with virtual ones. “Server Huggers” (found in IT and in the business) need time to come on board and support the initiative. You Will Most Likely Hit an I/O Bottleneck u You will most likely hit an IO Bottleneck Memory and CPU can be managed and allocated to keep up with most virtual server demands. NICs can be scaled and added as necessary. I/O, however, will most likely be a bottleneck and the limiting factor as to the number of virtual machines a host can manage. For example, this is particularly true in Exchange environments with consolidated storage and a high number of Black. Berry users. From a capacity planning standpoint, one Black. Berry Exchange user translates into four standard Exchange users on an I/O basis. Please see Page 36 for more details… Vendor Support May Be an Issue u Vendor Support may be an issue Although they are coming on quickly, not all vendors (especially smaller ones) are supporting their applications in a virtual environment. That being said, many enterprises we spoke to became reference clients for their software vendors once the virtual server implementation was complete. VMware currently has a number of initiatives underway to help gain software vendor support, one of which includes a group who works directly with ISVs. Of the top 70 ISVs, 75% now support VMware as a platform. Over time, vendor support will becomes less of an issue, as the vendors are recognizing virtualization’s prevalence in the market. Build in time to manage these issues with your vendor(s). www. infotech. com Impact Research 18
Common Pitfalls Heed the warnings of your peers (continued) u Implementation warnings continued Virtual Machine. Backups can be very large. Very Large u VM Image Backups Can Be Although a terrific feature, managing point-in-time snapshots of virtual machine images may not be worth the size requirement to make it beneficial. Beware of Vendor Licensing Arrangements u Beware of Licensing Costs Some vendors may still be inclined to charge based upon the total available power of the host system, as opposed to the apportioned pool of resources allotted to each virtual machine. Investigate licensing in a virtual environment when assessing applications for migration. Maintenance Contracts Could Bebe in jeopardy u Maintenance Contracts could in Jeopardy Much like vendor support, hardware hosting outsourcers may not support environments running virtual servers. Build in time to renegotiate contracts or find new service providers. www. infotech. com Impact Research 19
Common Pitfalls Heed the warnings of your peers (continued) u Post-implementation considerations Increased Pay/Incentives for IT Staff’s New Skill Set u Increased pay for staff due to skill sets Once staff is trained in the implementation and maintenance of a virtual server environment, think about financial incentives since they will now possess a hot skill set. This could be akin to what happened with ERP skills in their prime, although not of the same magnitude. While this shouldn’t be a grave concern, include it in the business case as necessary. Implement Appropriate Access Controls to Server Management Console u Appropriate Security Considerations Server management is made much easier through the implementation of a virtual server farm. However, this ease of management also requires appropriate controls and access permissions given the larger number of servers and scope of features that are being managed from one “pane”. Give few people access, and restrict features where feasible. www. infotech. com Impact Research The IT administrators at one client location would failover instances between physical hosts many times a day because “they could”. 20
Making the Virtualization Business Case www. infotech. com Impact Research 21
Business Case Areas The business case should be based primarily on hard cost savings u Virtualization has the benefit of having real cost savings; use these instead of quantifying other intangible benefits or cost avoidance items that may obfuscate the real cost-saving benefits which exist. u The preliminary business case can have a tolerance of +/- 25%, but be sure that all areas are addressed, as per the table below: Ongoing One-time Cost Savings • Facilities: Cooling, Power and Rack Space • Hardware Acquisition • Incremental Hardware Acquisition Cost Avoidance • Disaster Recovery and Business Continuity • Labor and Maintenance Savings • Host Machine Hardware Costs • Virtualization Software Maintenance • Virtualization Software • Staff Training (Turnover) • Purchase of the SAN (if required) • Salary Increases (In-demand Skills) • Staff Training and Consultancy Costs www. infotech. com Impact Research 22
Hardware Acquisition Cost Savings Hardware acquisition savings are the main force driving the virtualization business case Break-even Is 3 Virtual Servers per Host Machine u Depending on the approach taken, host machines for virtualization will either be purchased new, or existing high-end servers can be reused: • New server acquisition: the break-even point is collapsing three servers onto one virtual machine host assuming a two-way server configuration. • Reusing a server: The server can be reused in an other part of the business OR a host purchase is not required, saving an additional $3, 000 -$6, 000 and reducing break-even to a two-server consolidation. • This does NOT include the purchase of a SAN if one is not available in your environment. General Rule Virtual Server Break-Even $9000 -$18000 $6000 -$12000 $3000 -$6000 15 Servers Needed to Cover Implementation Costs u Including testing, labor costs, training and other implementation costs, the break-even point is at 15 physical servers or more. Please see Page 41 Blades Are Viable Options As Well for more details… u Blade servers are complementary to virtualization as they provide a hardware-based redundancy. 0 Cost of New Server $3 -6 K Acquiring New Servers Cost of Virtual Software $3 -6 K Cost of Virtual Host Hardware $6 -12 K Virtualizing u The Total Cost of Ownership (TCO) of an eight-server blade environment is now 21% less than its equivalent rack-mount server installation. www. infotech. com Impact Research 23
The more servers that can be consolidated onto the hosts, the better the savings (one-time and ongoing) 70% 10: 1 Consolidation Ratio A professional services client consolidated seven servers on each host z machine, realizing hardware savings of over 60% 67% 9: 1 63% 8: 1 57% 7: 1 50% 6: 1 40% 5: 1 25% 4: 1 3: 1 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% Average % Hardware Cost Savings www. infotech. com Impact Research 24
Facilities Cost Savings If deemed material, facilities savings and cost avoidance are a key driving factor u Facilities savings come from two key areas: Power Savings • Power: server power and cooling Savings in power consumption come from a reduction in the number of physical servers and reduction in the air conditioning needs. Rack Space Availability • Rack space Whether through simple virtualization or a combination of blade servers and virtualization, large amounts of rack space or raised-floor space are freed with server consolidation. u More than half of the clients interviewed state that shrinking the footprint was a driving force behind virtualization adoption. u However, this driver is less applicable to enterprises with unlimited space and large amounts of power in production, such as manufacturing companies. Bennett Jones, a large Canadian law firm, avoided the cost of building a new facility through virtualization. Graycon, a mid-sized professional services firm, emptied 40% of its rack space, and plans to have 60% emptied once implementation is complete. Virtualization can help to cut power consumption in half, while freeing up considerable rack space in the server room. www. infotech. com Impact Research 25
Cost Avoidance in Disaster Recovery / Business Continuity Enable high availability and DR u Implementation of virtualization automatically enables some level of high availability and business continuity. Cost avoidance areas include: Heterogeneous Hardware: No need for homogenous servers to achieve high availability/failover. Redundancy can be had without the need for “two of everything”. Hardware Sharing: Whereas applications, in the past, may not have coexisted within an environment, hardware can now be shared amongst competing applications as each will have its own dedicated environment. This enables failover to any available physical server. Failover without User Impact Failover without user Impact: Products like VMware’s VMotion allow for uninterrupted user experiences while moving instances between physical servers. Coorstek, a large manufacturing company, was able to cut its annual DR costs by over 35% because of virtualization. Backup and Restore – Snapshot Capability Backup/Restore: Snapshot Capability: Point-in-time snapshots can be taken of an entire virtual machine image, minimizing the amount of time necessary to return an application to a particular state. Note here that applications that rely upon data on a SAN would only have the local storage contents restored. Virtualization is software – hardware redundancy is still required. www. infotech. com Impact Research 26
Labour and Maintenance Savings Introduce efficient processes for server management More Efficient Processes u u Eliminates delay related to new server delivery, set up, and takedown. u Reduces development time by enabling the creation or refresh of development and testing environments in minutes. u Reduces time required to manage server farm by leveraging virtual machine management tools that allow for single-pane control of multiple servers. Repurpose Employees u u Efficiency leads to the ability to repurpose IT staff and get other initiatives underway which otherwise would have been postponed or incurred costs of contractor resources. Many companies interviewed report a drop in new hires despite growth of the business and IT environment. Clients stated that they are able to keep IT staff constant, despite growth in their business of up to 50%. Reduce Costcost of hardware maintenance u Reduce of Hardware Maintenance u Fewer physical machines can mean a tangible reduction in hardware maintenance contracts with server, or third-party, vendors Better staff utilization through virtualization allows more efficient management of the IT department, repurposing of employees, and meeting of business growth demands. www. infotech. com Impact Research 27
Low Benefit Areas Don’t look here for savings… License Costs u License costs u The only benefit noticed was with clients who used the OS-hosted Microsoft solution, which allows for four Microsoft servers to be run in virtual instances under one license. u Otherwise the number of OS licenses remains constant. Insurance Cost Reduction Insurance costs reduction u No noticeable benefits realized yet, but this may be something to watch for in the future. u FTE reductions FTE Reductions u In most cases virtualization is not significant enough to justify head-count reduction u www. infotech. com Impact Research 28
Necessary Expenditures Ensure the following key costs are included in the project budget The cost side of the business case should contain the following items: Please see Page 41 for more details… Hardware u u Acquisition of new servers to host virtual machines. • Multi processor servers or blade servers. The purchase of a SAN. Please see Page 42 for more details… Software u Virtualization software server license. Please see Page 43 for more details… Resources u u u Training. Consultancy costs. Additional remuneration for virtualization-trained IT staff. Many companies interviewed stated that they were able to use existing hardware as host machines, or repurpose existing hardware in other departments To achieve maximum benefit from the implementation, it is highly recommended that a SAN be included in the project budget. www. infotech. com Impact Research 29
Creating the Plan www. infotech. com Impact Research 30
Twelve critical planning steps for virtualization implementation u Based on successful cross-industry implementations, below is a compiled leading-practices guide to moving forward with virtualization. Step 1: Determining the Approach Step 7: Resourcin g www. infotech. com Step 2: Developin g the Business Case Step 3: Gaining Buy-In Step 4: Capacity Planning & Benchmar king Step 5: Hardware Selection Step 6: Software Selection Step 8: Applicatio n Sequencin g Step 9: Testing Step 10: Centralize/ Consolidat e Step 11: Migrate Step 12: Monitoring & Expansion Impact Research 31
Step 1: Determining the Implementation Approach Organization and project characteristics will determine the implementation approach IT Autonomy HIGH LOW High autonomy is characterized by: ü Executive trust in IT decision-making, resulting in a less formal approval process ü Relatively high risk tolerance within the IT group ü Consolidated IT budget versus departmental IT budgets Low autonomy is characterized by: ü Need for a formal business case and approval process ü Focus on hard cost savings and benefits as a result of C-level understanding of IT ü Non-IT executive involvement and responsibility for IT projects Scope of Implementation HIGH LOW HIGH ü Scope covers most systems ü Project goals are to maximize cost savings ü IT-specific systems being affected (e. g. internal help desk, DNS servers, Web servers, etc. ) Run as an internal IT project, and keep business and executives informed. Start small as an internal IT project. Learn from mistakes, and become more formalized with time. www. infotech. com LOW ü Scope covers many critical systems ü Many lines of business affected ü Minimizing overall upfront costs and risks ü Mostly development and testing systems are tackled Involve the business, and understand the full implications and Return on Investment (ROI). Consider consultants. Be transparent in your plans. Gain executive approval and support. Impact Research 32
Step 1: Determining the Implementation Approach Type of implementation is influenced by company size Scope of Implementation High Small Companies Large Companies Mid-Sized Companies Enterprise Companies Low IT Autonomy High Small and mid-sized companies are more likely to virtualize a higher percentage of their servers and execute their projects as an IT-only Initiative. www. infotech. com Impact Research 33
Step 2: Developing the Business Case Key components of the business case 1. Assess Fit: Assess whether the enterprise is an ideal candidate for virtualization based on fit indicators and benefit drivers described in this document and accompanying evaluation tool. 2. Determine Scope: Determine what servers and applications are candidates for migration. 3. Evaluate Alternatives: Make sure key categories, listed in the slides to follow, are covered: software and hardware selection (including the consideration of blades), the number of servers to consolidate initially, the implementation of a SAN, and training/external assistance. 4. Estimate Expenditures: Create a project budget by first identifying the necessary expenditures involved with the implementation. 5. Estimate One-Time and Ongoing Savings: Quantify the cost savings and consider the savings in cost avoidance that would be realized if they are tangible and recognized for the enterprise. 6. Estimate ROI: Determine the estimated ROI based on the above factors to be used for executive approval. Business cases typically have a +/- of 25, so make sure to caveat the estimates. 7. Develop Implementation Plan: Develop a timeline with key milestones for testing, training, implementation, and the date which the enterprise plans to revisit and expand the deployment. While the depth of the business case will vary, it is important to consider all of the mentioned components prior to implementation. www. infotech. com Impact Research 34
Step 2: Developing the Business Case Considerations that drive duration of implementation Drivers Number of Servers Being Migrated General Rule Duration of Virtualization Implementation Impacts As a General Rule, based on our analysis of various companies who have implemented from 15 to 200+ servers, a predictable correlation exists between the number of servers and duration. Staffing and resourcing implications are discussed in Step 7. Please see Page Total Duration = 2 Months + 1 Day per Migrated Server 43 for more details… Size of Organization Generally, larger organizations have more formalized processes in place, which means acceptance testing and sign-offs will extend duration. Criticality of Applications Being Migrated The more critical the application, the more time should be set aside for testing. Consider regression, user acceptance, and performance testing scenarios. www. infotech. com Impact Research 35
Step 2: Developing the Business Case Considerations that drive duration of implementation (continued) Drivers Application Vendor Support Impacts If the application vendor won’t support a virtual implementation, don’t think all is lost. In most cases we found the vendors came around with some proof-of-concept work and, in the end, used the account as a reference. Build in time for this “negotiation”. Almost all of interviewees stated that the project was met on time and was of anticipated complexity. www. infotech. com Impact Research 36
Step 3: Gaining Organizational Buy-In Three key points to help gaining organization buy-in u Depending on the approach that is most suited for the enterprise, deep buy-in may not be required, especially if the implementation is being run as in internal IT project. Appoint a Team Lead LEAD u APPOINT A TEAM A formal team lead acts as the key contact for issues, questions, and communication distribution. Having a point-person helps to alleviate user concerns by matching a face to the project. Bring the “Server Huggers” on. HUGGERS u CONVINCE THE SERVER Board Overcome initial misconceptions about the technology through education. A proof-of-concept in a very effective way of getting quick on-the-job training for the IT staff and addressing some users concerns. For executives, a well-positioned business case (with adequate contingency) is the best way to present the message as it is a compelling technology in that respect. A few successful clients have had great success in putting together a formal presentation, in laymen’s terms, that describes the technology and benefits. Roll It into an INTO AN APPLICATION RELEASE u ROLL IT Application Release As an alternative to a pure infrastructure project, a number of companies interviewed were successful in launching virtualization as part of a broader application release initiative (i. e. incremental servers). www. infotech. com Impact Research 37
Step 4: Capacity Planning and Benchmarking Four-step process to determining capacity 1 Identify application servers that are candidates for migration: u u 2 Determine the average utilization rate, and make note of peak utilization rates. Ideally, a month’s worth of data would represent an adequate sample. Choose the type of server that will host the virtual machines: u u u For most implementations, a two-way server is the minimum that should be considered given Hypervisor or OS overhead. For higher ratio consolidation effort machines, four-ways deliver the best value. Refer to the chart on the next page for virtual machine-to-processor ratios of successful implementations. One-way Typically Underpowered www. infotech. com Two-way Ideal for Low. Utilization Application Servers Impact Research Four-way Best Value for High Consolidation Ratios A large biotech firm hired consultants for the explicit purpose of creating a capacity plan to feed into the business case and implementation plan. Eight-way Price Prohibitive 38
Step 4: Capacity Planning and Benchmarking A good guideline is to consider three to four virtual machines per processor u Most companies have found a good mix of performance and consolidation value around three to four virtual machines per processor. u To be more accurate in planning, you can consider the number of virtual machines per processor core. Low 0 1 High Median 2 3 4 5 6 7 8 9 10 Number of Virtual Machines per Processor www. infotech. com Impact Research 39
Step 4: Capacity Planning and Benchmarking Four-step process for determining capacity (continued) 3 Set the target utilization rate for the host server considering failover capacity u u 4 Target utilization on host machines should be no more than 60% if the plan is to fail instances to it. About 80% should be the maximum utilization on a host server in order to allow for adequate resource balancing for spikes in processing need. Performance test to account for memory, NIC, and I/O constraints. u Be wary of any hard-stated recommendations of consolidation ratios or host server capacities. www. infotech. com Impact Research Basin Power Electric Company went from having an average server utilization rate of 3% prior to virtualization, to host machine utilization of 50% with a 17: 1 consolidation ratio. 40
Step 5: Hardware Selection Blade servers are complementary to virtualization software and a viable choice u While virtualization software provides redundancy at the OS level, blade servers can perform a similar function at a hardware level. u Consider the merits of both when making a decision. Blade Server Considerations Multi-processor Server Considerations PROS • Can host large numbers of virtual machines in fourway and eight-way configurations • Existing servers can be upgraded to act as host machines PROS • Have very small footprint and provide good scalability (limited by chassis) • Provide some level of hardware redundancy between blades CONS • Can be costly to outfit with internal hardware redundancy CONS • Run very hot for their size (dependent on number of blades of course) Although not mandatory, a SAN will enable many of the benefits of a virtualization implementation. Make sure to include this expense in your calculations. www. infotech. com Impact Research 41
Step 6: Software Selection VMware is the current preferred product and market leader Vendors Price u Unsupported version is free u Free u Supported Enterprise version u Hosted solution is free u Hypervisor 2 -Socket license is $750 ranges from $1, 000 to $5, 750 Need for Host OS? u No – includes Hypervisor Supported Guest OSs u Linux, BSD in free version u Yes – Windows u Free version requires Windows or Linux u Licensed version includes Hypervisor u Windows, most Linux u Windows, MS-DOS, Linux, Novell, FBSD, Sun Solaris u Windows, Linux, BSD, and Solaris in Enterprise version Benefits u Open-Source product offers very reasonably prices toolset u Install up to four copies of a server OS on one license u Mature tools for resource management and business continuity Our interviews showed that the majority of companies are using VMware products, with a few running Microsoft Virtual Server. www. infotech. com Impact Research 42
Step 7: Resourcing Start by estimating the total number of resources required u Determine resource and staffing needs as this has potential impact on three types of costs: • Training costs • Costs of backfilling seconded staff • Consultant costs Estimate Staffing u Estimate Staff Needs u Determine how many staff should be involved to execute the implementation within the duration desired. A general rule for calculating staff requirements based on work effort appears on the right. Total Work Effort (in Person Weeks) = (2. 9 x # of Servers) -31 Staff Requirements = Total Work Effort (in Person Weeks) # of Weeks of Desired Duration are taken into consideration. Determine Availability and Capabilities of Staff u Determine Internal Availability and Capabilities u Next, decide if the skills sets required are available and the resources with those skills have capacity to take on additional tasks, or be seconded to the project full-time. Impact Research Number of Servers 15 Duration in Weeks 4 15 8 1. 6 30 u Ensure that project complexity and organizational risks www. infotech. com General Rule Work Effort Required for Implementation 8 7. 0 30 16 3. 5 60 16 8. 9 60 26 5. 5 Staff (FTE) 3. 1 43
Step 7: Resourcing Company size is a good indicator of what type of training or external assistance you should investigate u Once staffing needs, resource availability, and capabilities have been identified, consider the mix of training and external consultants that you will leverage. u Our interviews led us to the conclusion that Small Companies u Small Companies Given a smaller staff with varied duties, small companies tended to allow their staff to learn on the job while executing the project. Medium-Sized Companies u Medium-sized Companies These companies would send one or two resources to training to become the trainers for the rest of the staff. Large Companies u Large Companies In most cases, given the transparent nature of the project, and the various parties involved, consultants were brought in to provide guidance and ensure an expedient implementation. www. infotech. com Impact Research High Consultants Cost of Resourcing staffing mix and training are strongly correlated to the size of company, not necessarily the size of the implementation: External training On-thejob training Low Small Medium Large Size of Company 44
Step 7: Resourcing Key considerations when evaluating training and external consultant alternatives Consulting Costing Questionnaire Staff Training Costing Questionnaire 1 Do my staff have the understanding to build these new skills? 2 Do I need to backfill staff on training? If not, consider consultants and/or new hires. 3 1 Do I need consultants for point expertise to complement your team? 10 -20% of total project time is probably adequate. If so, what is the cost? 2 Do I need the consultants to augment my team under the assumption I can’t meet my deadlines with current staff? 3 4 Do I have high turnover? Should I include a salary premium? Do I need consultants to run the project? If so, consider the recurrence of training costs. If training, consider that these staff are now more valuable in the market No more than 80%. Ensure a project manager comes with the consultant staff. Maximum 80% staffing of consultants is recommended. internal staff should be involved to some extent. Recommended to keep the mix around 50%. The larger your implementation, the proportionally less relevant the training and consultancy costs. www. infotech. com Impact Research 45
Step 8: Application Sequencing Virtualize low-utilization applications if the project is an internal IT initiative LOW-UTILIZATION INTERNAL IT APPLICATIONS CRITICAL APPLICATIONS Migrate critical applications first when: Migrate low-utilization internal IT apps first when: u A business case has been created and formal approval process completed. In this scenario, business representatives are involved at the appropriate levels to gain buy-in for a critical application. u The project is being run as an under-the-radar internal IT project with goals to minimize risk. u Cost savings is the main concern. u Gaining buy-in and building faith in the technology u Improving DR capabilities is top concern. u Consultants have been brought on board to guide or execute the implementation. Their experience should be leveraged towards critical application migration. are key goals, with plans to move to more critical systems later. Note: Not every service can be virtualized and some services require special attention in order for virtualization to work. Investigate vendor licensing in a virtual environment first – some vendors may still be inclined to charge based upon the total available power of the host system, as opposed to the apportioned pool of resources allotted to each virtual machine. www. infotech. com Impact Research 46
Step 9: Testing Step 10: Centralize and Consolidate Given the nature of this technology, adequate stability and performance testing is a must u Two types of testing are critical to the success of a virtualization implementation: u Virtualization is a consolidation exercise and, as such, centralization of servers is key to achieving favorable consolidation ratios. Regression/Stability Testing u Regression/ Stability Testing The testing of an application’s behavior in a virtual environment cannot be predicted and must be tested thoroughly before production launch Centralize First u The majority of our interviewed companies were already centralized, or centralized a portion of their infrastructure prior to virtualization. Performance Testing / Tuning u Performance Testing Even with a detailed capacity plan, some performance testing and tuning must occur in the testing environment. Resource allocation and load balancing features must be configured and tested to handle peaks. Unless you have a Business Case Not to u We did find, however, that mini-business cases can be created at a branch level given a critical mass of servers at each location, and a justification (either bandwidth or staff location) for not moving servers away. Info-Tech recommends purchasing a SAN when implementing virtualization. www. infotech. com Impact Research 47
Step 11: Migrate Roughly 85% of your time should be spent on implementation steps u Percentages listed below are rough guides for time allocation across the steps. u Pre-implementation activities will take longer in larger organizations than smaller ones. u The key to success is getting into the implementation steps quickly and working out the issues in a testing cycle. u Once testing is complete, the migration may take the same or less time to complete. Pre-Implementation Approach Acquisitions Implementation Steps Staff Training Testing Results Centralize and Consolidate Partnerships Business Case Development Time to Implement Expanded Services Organizational Buy-in Migrate Average Implementation Takes Seven Months Status Planning Capacity Quo 5% 40% Exit / Sell Hardware/Software Selection 15% www. infotech. com Impact Research 48
Step 12: Monitoring and Expansion Virtualization implementation is an ongoing process u Once the initial implementation is complete, key processes and plans need to be put into place: Create a Roadmap of Application Servers to Migrate u Roadmap of Applications to Migrate To continue the benefits stream, a roadmap of future application servers to migrate should be created (assuming the initial implementation migrated a subset of servers). Establish Process for Requests u Process for Server Instance Requests Virtual server sprawl is as much a problem as physical server sprawl. A request process should be put into place to manage server instances. Continue Ongoing Capacity Monitoring and Planning u On-Going Capacity Planning As additional instances are migrated and as application dynamics change, constant monitoring of resource pools is required to ensure performance targets are being met. This is especially important as physical hosts are targets for DR of failover instances. Begin Cross-training Immediately u Cross-Training System administrators will now have control over a larger pool of server instances. To ensure succession planning in the event of turnover or absence, a knowledge transfer and cross-training plan should be put into place post-implementation. Collect Original Business. Metrics u Collect Business Case Metrics To maintain faith in IT’s abilities to deliver value, close the loop and collect post-implementation metrics to validate the original business case. www. infotech. com Impact Research 49
Step 12: Monitoring and Expansion Implementing virtualization can be an entry into utility computing u Despite virtualization being an infrastructure initiative, there are process and strategy ramifications on how IT can/should operate with the new flexibility of virtualization. Transition to Utility Computing u Chargebacks and Transition and Chargebacks Virtualization inherently lends itself to utility computing, where each virtual instance, the memory, CPU, and disk it uses are procured as required and charged for. Chargeback models, if a cultural fit for your enterprise, are complementary to virtualization in that measurement of chargeback metrics is made easier as is the provisioning of capacity to meet various Service Level Agreements (SLAs). www. infotech. com Impact Research 50
Скачать презентацию The Business Case in Virtualization www infotech com
b774a66b1f1edb188b072034b20c901e.ppt