VMware has clearly become the standard solution for Wintel virtualization, and offers the potential to significantly reduce cost and complexity. While in theory eliminating as many Windows-based servers as possible would maximize efficiency, decisions on what to virtualize aren't always straightforward. The importance, nature, cost, and business requirements related to applications need to be considered. Also, while many of the best practices for managing the data center carry over from the physical to the virtualized world, virtualization introduces new challenges related to culture, resource requirements and processes. All of these factors need to be carefully considered in the context of an IT resource optimization strategy.
This article highlights some of the more important best practices that must be in place when moving to a virtualized infrastructure.
Leverage a portfolio management discipline to determine what should be consolidated or virtualized and how.
Portfolio management enables IT and the business units to make informed decisions that consider the needs of the business and the importance or value of the applications. In the context of virtualization, portfolio management is key to helping prioritize what can and should be virtualized. Regulatory concerns, mission criticality of applications, business needs and cost structures can all have a significant impact on virtualization decisions and are highlighted by portfolio management processes. By taking a broader view of applications and systems, you can also avoid getting trapped into a virtualization-only consolidation strategy and explore alternatives that may offer new and greater savings or efficiency while minimizing risk. Portfolios also provide a basis upon which to build a business case for consolidation and virtualization by accurately tracking and allocating costs and potential savings.
Implement or redesign chargeback models.
Chargeback – still an area of contention for many organizations – can be more of a friend than foe in a virtualized environment. Politics are a very real danger to virtualization initiatives. Business units can wield their significant influence to quickly squash an initiative that's seen to place their applications in any way at risk. Accurate, transparent, equitable chargeback clearly demonstrates the business case for virtualization. A simple allocation model that distributes costs according to a pre-determined percentage, or one that does not differentiate between logical servers and physical servers, will not motivate the business units to support virtualization. Tools that monitor utilization and facilitate accurate cost allocation on that basis make the cost savings transparent to business users and help diffuse political debates. These can support chargeback systems that give business units a visible improvement in their bottom line as the savings from virtualization are realized.
Develop the Capacity Planning function.
Previously, capacity planning was a concern primarily in the mainframe world, where multiple users and applications shared resources. Virtualization introduces a cultural change for mid-range system managers who traditionally resolved the rare issue of workload conflicts by simply purchasing a new box. Capacity planning in the virtualized data center is an art of balancing risk and maximizing utilization. New tools exist to help plan and monitor capacity within virtualized environments, but teams also require the skills and processes surrounding those tools to do the job effectively. During the planning phase, organizations should seek out tools that accurately model workloads, including overheads created by VMware's virtual machines to optimize placements. Once in production, key features of VMware such as Vmotion VMotion and Distributed Resource Scheduler (DRS) can enable dynamic workload balancing by migrating VMs to new targets without interrupting operations. While relatively new, VmotionVMotion is rapidly becoming popular as a cheaper and comparable alternative to managing traditional clusters. A virtualization planning solution by CiRBA maps out areas of Vmotion VMotion capability within your environment and offers integration to VMware's DRS, automatically populating rules that govern where servers can be moved to avoid potential conflicts and reduce risk.
Prepare to invest in training.
Provisioning and staffing processes need to accommodate new technologies. Significant changes follow in provisioning the new servers. While organizations will no longer require the hands-on box gurus, a more complex administrative role will be required. A skills shortage associated with VMware-based virtualization is becoming a growing issue. The popularity of the tool has created high demand for experienced resources – and since the technology is so new, "experience" is hard to come by. This not only drives up employee costs, but means organizations are often left training their own staff and venturing into this new territory without the benefit of experience. In addition, ongoing training is required to keep managers and technical staff current with a rapidly evolving product suite and new complementary solutions.
Implement new processes to prevent virtual sprawl.
In the physical server world, purchasing departments serve as gatekeepers to help stem server sprawl. While the proliferation of Wintel servers may suggest they weren't that effective, today's virtualized data centers can be essentially free of any monitor. In production environments the individuals creating virtual machines are typically also responsible for managing them. This direct relationship means this group is less likely to freely create new VMs and increase complexity. On the other hand, application development (AD) teams may have no such motivation. AD teams might find it useful to create new servers for testing and staging, but just as their physical counterparts were underutilized and abandoned, so too the virtual servers will be. IT teams will be faced with a real cost in terms of increased complexity and infrastructure management time if this becomes the norm. Best practices dictate that new policies be defined to include justification, parameters, and documentation for creating each virtual machine. Implementing tools that track systems and virtual machines through audits can help keep VM proliferation under control.
Adding a virtualized infrastructure to a data center can have far reaching impact into staffing requirements, management processes and policies. Organizations that are successful will be those that plan in advance for changing requirements and develop new best practices to ensure that the goals of reduced cost, improved responsiveness and better reliability can actually be achieved.
About the author: Scott Feuless is a senior consultant for Compass America, a global management consulting firm specializing in business and IT performance improvement for Fortune 1000 organizations.