Trends for successful datacenters: virtualization, cloud, storage

When the New York Times criticized datacenters for their wastefulness, it wasn’t just a little off-target. It entirely missed, for example, the Open Data Center Alliance‘s prediction “that the overall power savings [emphasis added] from cloud adoption could add up to approximately 45GW by 2015 … enough … to power up to 15 million homes.”

Or maybe the more important point is to drop the pursuit of silly metrology aimed at connection with lay readers (how many truly have an accurate feel for the difference between a million homes, and a thousand?), and focus on sober comparisons of real-world experience. As Robert McMillan of Wired and others, include the Real User Monitoring blog, have repeatedly described, there’s a lot we simply don’t know about how datacenters are actually run. A survey from this summer recently reported in Data Center Journal though, gives at least a few tantalizing hints:

Win at virtualization

Virtualization has been a win, and has plenty of potential for more gains in the future. Virtualized storage still is deployed in only “21 percent of less advanced environments.” Perhaps as important as tangible immediate energy savings, complementary human efficiency also grows: the survey found that more-experienced datacenters run with 8.2 virtual machines (VMs) per physical server, compared to a ratio of 4.5 “for less virtualized data centers.”

Virtualization structured as “private cloud” is the strong new trend the survey unearthed: “more than 75 percent of all participants highlighted private cloud as their primary data center growth strategy”. The Data Center Journal article puts it well:

Cloud architectures appear to solve three fundamental challenges facing today’s data centers:

  • Delivering elastic services to accommodate a wide range of demand fluctuations
  • Automatically provisioning and maintaining a common pool of compute, network and storage resources rather than the current static provisioning model
  • Extending limited capacity headroom

All these results bring us back to the Real User Monitoring blog’s persistent themes: how do we measure performance in meaningful business terms? What decisions can we make or projects undertake to improve operations? Over the next couple of weeks, we’ll look at a few of the best ideas in platform-as-a-service (PaaS) and other cloud approaches, end-user oriented application performance management (APM), and coordination of technical opportunities and business goals.