The Server Consolidation Bottleneck

It would be unoriginal for me to submit some sort of apology for the delay between my posts, so I won't offer one. I'm unsure who would care if I neglected to include one anyhow. ^_^

One of the subjects that I have been doing a lot of business research into of late is one of the current IT industry darlings, virtualization. I have a lot of personal and professional interest in learning more on this topic because its impact on how traditional proprietary software companies license their software and because the flexibility and efficiency opportunities that it presents businesses with a significant IT infrastructure.

Like a lot of new technologies, I think there is a gap between the promise of virtualization and the reality of what can be delivered with the current state of the technology. One key example of this gap is the best-articulated business value of virtualization, server consolidation. I recently attended a seminar held locally by VMware where the typical server consolidation ratio for VMware customers was cited as being between 8 and 12 to 1 with some customers achieving a 30 to 1 server consolidation ratio. A 30 to 1, 12 to 1, or even 8 to 1 consolidation in server hardware is remarkable to consider in any case, and would likely get the attention of anyone seeking to reduce the acquisition, deployment, support, power, cooling, and maintenance costs of servers within an IT organization.

So what's the gap here? Basically, that degree of server consolidation assumes that the applications running on the guest machines at most, use the network infrequently or not at all. A recent Wall Street Journal article (Real Virtualization Battle Looms, 2/26/2008) highlighted the "communications bottleneck between server systems that has made it impractical to use virtualization for some of the most demanding applications, such as large databases". In essence, this says that forcing applications requiring significant use of the network to share the network I/O resources of a single server with other operating systems and applications isn't practical. The WSJ article does note that there are some organizations looking at solving the problem of managing network I/O for virtualized environments, but at the same time, if an organization has many applications already running near capacity on a robust hardware configurations, it seems unlikely that virtualization alone will offer them much in the way of a consolidation benefit since increasing network bandwidth isn't easy or cheap.

I do think that the future of virtualization is bright and offers a great deal of promise for the future of IT organizations of all shapes and sizes. Until the practical challenges of consolidating network I/O-heavy applications to a significant enough degree to experience real cost savings is solved, however, the full scope of virtualization's server consolidation benefits will remain untapped.

virtualization server+consolidation VMware application+performance network+efficiency