Market Perspective: How Organizations are Delivering Value With Virtualization

1. Market Perspective:How Organizations are Delivering ValueWith Virtualization Shawn P. McCarthy June 21, 2007

2. Discussion Topics • Overview of virtualization (strategies & definitions) • Market relevance • Industry examples • Common misconceptions • Assessing the cost/benefit of virtualization approaches • Operational impacts of virtualization • Virtualization’s place in current/emerging computing trends • Security and IPv6 transition considerations • Return on investment (ROI) considerations • Challenges • Summary • Q&A

3. What is Virtualization? • The abstraction of computer resources • Effectively presenting machines, databases, IT services and interfaces in a way that allows applications, other systems, and end users to transparently interact with those resources • Associated functions • Make a single physical resource appear to function as multiple logical resources • Make multiple physical resources function as a single logical resource

4. The “New” Virtualization • Historically the term meant: • Creating a system within a system • Partitioning hardware into multiple virtual machines • Simulating one operating system or application within another system • Today virtualization means: • Merging analogous system resources into carefully designed resource groups • Allowing these to be ramped up or down as demand dictates

5. How Is a Virtual System Built? • A virtual system doesn't necessarily need to be constructed as a blade system or a centralized server farm • At the start, virtualization can be done over an existing network and via existing machines • An application running in a virtual environment must be constructed to interact with specific registry entries, files, and components • Properly designed, this environment fundamentally operates as a separate layer between the application and the server operating system Applications can even reside within the virtualized environment. Java applications are an example of this.

6. Virtualization Terms and Definitions Basic Terms • Host: A computer, on which the emulator/virtualizer software runs. • Guest: The system you are emulating on the host computer. Often considered a virtual machine or VM for short. • Hyperviser/Virtualizer: Software that does (Full/Para) virtualization and monitors one or multiple virtual machines.

7. Virtualization Terms and Definitions (Cont.) Types of Virtualization • Emulation: Provides conversion of commands, or instructions via software. Complex commands can have a high overhead for proper conversion. • Examples: DOSbox, Bochs, Qemu, Dynamips/Dynagen, ePSXe. • Full Virtualization: Allows execution of some commands on a CPU without translation. Can be done with solutions that provide OS-level integration for many commands. • Examples: VMware, Qemu +accelerators (KVM, QVM, KQemu), VirtualPC, VirtualBox (and Xen has limited functionality here). The Linux kernel has KVM in this area. • Para-Virtualization: Similar to full, but modifies both host and guest operating-system kernel. • Examples: Xen, User-Mode Linux (UML), VMware6. • Operating-System Level Virtualization (Containerization): Provides the tightest integration between host and guest systems, by providing a single, unified, heavily-modified kernel. • Examples: Linux vserver, OpenVZ, Virtuozzo, FreeVPS

8. Virtualization Examples • Storage • Logical volume management tools can combine multiple disks across a network into one large logical disk • Network bandwidth • Technologies such as redundant array of independent network (RAIN) interfaces can merge several network access points • Allows them to function (when needed) as a single, higher-bandwidth link • Applications • Software no longer needs to be “installed” on the machine where it will run • Runtime environment can now isolate the application from hardware resources and the APIs of an underlying operating system • The most effective application virtualization systems can dynamically assemble applications in real-time and load balance across multiple processors or machines

9. More Virtualization Examples • Scaleable server partitions • Dividing a computer into several independent virtual computers • System logic adjusts dedicated processing power as needed by individual applications • Most DBMS products today allow partitioning of large tables, the database can be scaled out across a cluster of separate database servers • Threaded processing • Multithreaded processing has been around for years • Virtualized systems allow better thread control to leverage all available processors or other attached systems

10. Consolidation is JustThe Starting Point Why Virtualization? • System consolidation isn't just about reducing numbers of machines and software licenses • It means developing solutions that move toward standardizing an organization's architecture • System managers face a challenge – managing their current systems while migrating toward eventual component standardization Incremental steps are a must!

11. Market Relevance

12. A Few Industry Best Practices • Assure that host hardware is on the virtualization software vendor's hardware compatibility list. • File management can get complicated on virtual machines. Virtualization software often uses cryptic file and directory names during the guest VM creation process. Don't be afraid to customize VM files to help identify host names and locations related to shared directory. • Enhance security by limiting host access and guest VM management. • Set performance goals. When loading-up a host server with VMs, 60% to 80% total utilization is a reasonable target for maintaining overall host and guest virtual machine performance and high ROI.

13. Government Virtualization Examples • IRS Applications Modernization Vision and Strategy • IRS will couple some components of its account management services to new releases of its taxpayer database, the Customer Account Data Engine. • Applications and components reside across multiple servers. • Miami-Dade County • Workload of about 100 servers has been migrated to three high-capacity servers. • Moving toward a more on-demand system in order to achieve this. They have been migrating to powerful servers, including a blade system. • State of Utah • State has moved heavily towards a service oriented architecture, with over 850 major systems & components. Allowing these to exist as discrete services that can be called from anywhere has allowed the state to break down previously existing silos.

14. Government IT Services:A Virtualization Approach • Some agencies are now developing new IT services nearly every month, making them available on their networks where they can be called by other applications. • Reducing complexity and forcing all systems to work in concert, some also are leaning toward system virtualization as a solution to their consolidation challenges.

15. Governments as a Proving Groundfor Virtualization • As managers work to consolidate applications across multiple departments and agencies… • A virtualization environment can employ grid scheduling or advanced threading techniques • Allows a process to be redirected at any time to a different resource • The process will continue to execute where it left off upon its arrival at the new system • Resources increasingly are hosted on groups of shared servers • Able to call data and services from multiple other machines or systems • Allows for easier threading and resource sharing

16. Governments as a Proving Ground for Virtualization (cont’d.) • The main drivers • Cost savings • Ease of administration • Reduced overhead • Improved security • Current government adaptation of virtualized environments • Often driven by IT systems integrators who are tasked by agencies to reduce their machine overhead or numbers of software licenses • Also being driven by builders of high-end servers and systems • Together, they support multi-threaded processing and embedded special technologies that enable service aggregation and management resources – with a consolidated, logical view

17. Common Misconceptions • Virtualization will save money by reducing personnel requirements • You will still have the same number of logical servers and staffers are still needed to manage every virtual machine instance. • Large organizations may see some staff reductions. • It is possible to save money on power and space costs. • Server consolidation saves money on software licensing • Today most vendors have developed advanced software to take advantage of virtualization, and it usually costs more money. More end users can also boost the cost of individual licenses. • Server virtualization boosts reliability • This depends on configuration, how machines are accessed, etc. In some cases, business continuity is an issue if you have a single point of failure.

18. Assessing the Cost/Benefit of Virtualization Approaches • Establish a baseline of current costs, including the cost of “doing nothing” • What will your organization need to change to move toward virtualization? • Hardware costs • Software costs (including any optional features) • Network improvements • IT services or staffing changes • What are the maintenance fees? • Don’t forget the licensing costs for the operating systems and applications that will be run within guest instance • What is the impact of adding CPUs? • Do multicore chips have higher per-CPU licensing costs for your software? • Government ROI can also include improved citizen services, response times and security

19. Longer-Term Savings #1Improved Data-sharing & Smaller Apps • As systems are consolidated, it’s easier to share data and certain processes across machines. • May be offered as discrete IT services rather than full-blown applications. Can be called upon by multiple other applications on the same machine or across the network.

20. Longer-Term Savings #2Speeding Development Time • Old development method • Separate development server (or network) with an application prototype • Full detached development of all parts of the application • Expensive, slow and unwieldy Once Virtualization is Deployed • New development method • Partition unused space on a server or blade system • Segregate the new application with the partition • Call existing external data, application components or IT services as needed

21. Operational Impacts of Virtualization • Is your data center equipped for a migration to virtualization? • Does your system support staff understand virtualized environments and are they trained in the solution you have selected? • Business continuity on virtualized systems must be carefully planned • What sort of domain name system configuration will you need? • How are you handling security certificates and system access?

22. Enterprise Virtualization • A full virtualization infrastructure might look something like this

23. Potential Risks Associatedwith Virtualization • Political challenges • Scope & command of a typical administrator is limited to a specific application or system • No immediate career advantage to migrating an application to a centralized server • Enterprise architecture challenges • Need a broad commitment for multi-application virtualization, data and resource sharing • Constituents may have issues with abandoning some systems • Funding challenges • Tough to find investment for migrations that may take a few years to mature • ROI models need to be proven

24. The Larger Virtualization Trends • Tying together different programs and data sources over a network • Using component services to construct order to build entirely new applications • New services can be added in a way that allows any application to take advantage of that service • Government offices often start with • Financial • Human resource • Supply chain management • Almost any process can be virtualized

25. The ROI Question • Need to establish a baseline of costs before beginning your migration • Organizations that can truly benefit from virtualization often achieve a full ROI in under two years • Organizations may find it difficult to achieve a positive ROI if: • Their IT operations are already running efficiently • Migrating to virtualization will require new servers, new applications, new network software/infrastructure and extensive training • In such cases, virtualization can still be a long-term goal via enterprise architecture rules for any new system

26. ROI Considerations • Software licensing costs • Current server costs, including power consumption • Bandwidth (Costs? Savings?) • Headcount • Service to citizens • Maintenance reductions • Paperwork reductions • Overtime reductions • Current and future reporting requirements • Other government migration requirements (IPv6, HSPD-12)

27. The Big Challenges • Quality of service • System response times • Engineers who are capable of realistic planning to offset the two challenges above • Should be able to develop a map of how and when resources are needed • Arranging for adequate processing power • Taking a componentized approach can actually speed services in many circumstances because the burden of executing tasks doesn’t fall on a single machine – but the system design must be correct

28. Other Challenges • Business objectives of both the government agency and its IT department must be clearly defined • Systems must be constructed in a way that allows for a virtualized and expandable solution • Agencies must be able to dedicate enough funding to start a true virtualization effort • Many IT budgets are mainly targeted at keeping existing systems running • Good first step: Establish a plan that allows you to approach virtualization in a series of steps expanding services and eventually processing power as you go

29. Next Steps • Create pools of similar resources such as servers, storage devices, possibly databases -- work to virtualize them across available machines. • Then start virtualizing resources at the application level. This may mean calling data and services from different locations, but managing the data as a single entity. • Next, start virtualization efforts across the whole enterprise. It's not about creating islands of virtualization, but rather re-architecting the whole IT infrastructure. • At some point virtualized resources can be offered externally, as a way of sharing data, processing power and services with other entities, such as other government agencies or commercial partners.

30. Once the Virtualization Effort is Underway – Start Promoting On-demand Services • Identify all services which exist on your network. Make sure you understand exactly what they do and how they may be of use to additional users. • Write adapters to wrap these services in a way that they can be located, called and used by other applications. Multiple services should be made available in a uniform way. • Install a common communications policy, so all systems and components can interact with each another. • Establish a registry so programmers can identify applications and use them. • Consider building a business rules engine that can run a composite application when called by a user.

31. Summary • Virtualization works because it helps organizations take advantage of all of their resources and processing power • Properly executed, an enterprise-level virtualization effort can help build a long-term IT architecture that can be leveraged in new ways • Such a transition only works if it is carefully planned and executed • While many of its parts can be commercial off the shelf products, some level of system customization will always be needed in a government virtualization solution

32. Questions? Comments? Shawn McCarthy smccarthy@government-insights.com