In deploying and managing virtualization, one of IT's most persistent problems is overcoming the political issues that come with major technological change. These issues often center on the fundamental conflict between business units and IT over which unit owns and pays for servers in a virtual data center.
In this tip, I'll identify some basic virtual systems management (VSM) disciplines, such as chargeback and capacity planning, that help eliminate business-unit wrangling and enable IT to develop an approach to future service delivery.
How virtualization has changed server allocation and ownership
When applications moved from multi-tenant environments like z/OS, VMS and AS/400 systems, business units and department managers effectively began buying and using their own dedicated physical systems. Each business application ran on a specific server in the data center, usually sporting a distinctive label to identiy the server's purpose. Each department requisitioned its own servers, paid for them out of its budget, had IT install them and their applications, and then owned them completely until they were decommissioned.
With virtualization, this is no longer true. Business units can no longer point to a physical server and say, "I own that server," because the workload may be migrated from one physical server to another at any time. Now they also share server resources with other departments and applications. Even though they paid for a server from their own budget, if they don't use all of it another department gets what it perceives as a "free ride" on its dime. Yet they're probably unaware that they, too, may be getting a "free ride" whenever they need additional server resources, have a server fixed or experience a catastrophic failure. All this generates many difficult and fractious discussions about who owns the servers and who should pay for them.
Settling server ownership issues with VSM disciplines
In order to deal with these discussions more effectively, it's worth taking the time to employ some basic virtual systems management disciplines, many of which provide lessons from the original multi-tenant environments. Some important approaches include the following:
- Chargeback. As applications moved off the mainframe, this discipline evaporated but has gained increased relevance with virtualization. Virtualization-aware chargeback systems provide highly accurate measurements of utilization by department, application, user or transaction, regardless of the complexity of shared systems and dynamic migration. Whether the resulting chargeback metrics are for reporting or used for actual cost recovery, they provide a solid foundation for any discussion over which unit uses virtual resources and how much it should pay for them. IBM Tivoli, PlateSpin, and Vizioncore all provide solid solutions for virtualization-aware chargeback.
- Resource measurement. Even without formalized chargeback, simply measuring resource utilization and providing reports to business units can keep you out of trouble. Reliable metrics of who uses which resources and how much they use, can prove to business leaders and budget owners that they've gotten what they've paid for and even show where resources and systems bought with another business unit's budget are being taken advantage of. They also provide the basis for better resource utilization and allow business owners to see that they can enjoy service-level agreement (SLA) -compliant performance, even while sharing a server with other business applications. HP, Hyperic and Vizioncore provide resource measurement tools specifically supporting virtual environments.
- Discovery and asset management. Virtual machine (VM) sprawl and other uncontrolled software deployment can quickly prompt server purchases to address unexpectedly high system utilization, fines or punitive costs because of license compliance breaches and unanticipated costs for new software licenses. These costs must be borne by IT or passed on to business owners. If IT automatically discovers, records and reports on the software each business unit runs, it knows quickly which business unit needs to pay for these systems and licenses. Simply providing consistent reporting curtails or reduces unnecessary hardware and software usage. CA, Tideway and Symantec all provide serviceable virtual asset discovery and management solutions.
- Capacity planning and management. Understanding resource requirements and compatibility of different workloads helps to predict future utilization rates, forecast future resource requirements and provide rational and predictable cost forecasts to the business. This can allow for a quick response to business requests with existing hardware, help to avoid a last-minute fight with budget owners over buying a new server and let you deploy new servers on a schedule and budget approved months ahead of time. BMC Software, CiRBA and PlateSpin all provide solutions for capacity management in virtual environments.
Applying VSM principles across environments
It's also important to take environments such as testing, acceptance, quality assurance (QA) and disaster recovery (DR) into account. Server ownership is not just an issue in production. For the business units that provide and/or approve IT budgets, every system matters. Monitor and measure them all.
A few advanced IT environments have addressed these problems more holistically by providing a so-called cloud-computing service to their internal users, with a pool of virtual servers that can be automatically provisioned and charged out on an ad hoc usage basis. Over time, all IT organizations should strive to achieve a level of dynamic, service-driven computing, and these fundamental VSM technologies are critical to this model. They also help IT address political issues immediately, provide a tactical solution to an immediate problem and position IT to deliver a strategic approach to future service delivery.
ABOUT THE AUTHOR: Andi Mann is a research director with the IT analyst firm Enterprise Management Associates (EMA). Andi has over 20 years of IT experience in both technical and management roles, working with enterprise systems and software on mainframes, midrange, servers, and desktops. Andi leads the EMA Systems Management research practice, with a personal focus on data center automation and virtualization. For more information, visit the EMA website