From VM Sprawl to VM Stall: Why Visibility Matters in 2011

Starting in 2009 and throughout 2010, I’ve been very focused on how the cloud will be managed. For successful Tier 1 enterprise apps to move to the cloud, we need visibility and control over workload configuration as well as performance—and we need both at levels beyond what many have achieved in the datacenter itself.

For a cloud deployment to deliver significant ROI, automation is critical. There’s no room for mad scrambles to diagnose a new performance problem in the dark, or to manually provision new workload instances. IT will need to automate the lifecycle of a complete workload—including compute, network, and storage resources—and therein lays the problem.

Even if an enterprise datacenter has as much elasticity in the network and storage layers as it has in its virtualized server layer, coordinating resource allocations, automating multi-tiered provisioning, managing resource contention and diagnosing tricky I/O performance problems across multiple technologies, operational silos, and business units quickly becomes overwhelming.

The first generation of virtual infrastructure management tools were all about visibility: what do I have, who’s using it, and how many more can I support? Provisioning and inventory automation were the most important IT tasks to automate. The watchword for this era was VM sprawl: how to rein it in, how to get our hands around it, and how to avoid it becoming a blocker to further virtualization.

Now, as virtualization has matured into a core datacenter and cloud operating platform, customers are less worried about sprawl (too much of a good thing can still be a good thing, after all) and are instead facing the threat of VM stall. We're exiting the first wave of VM-based consolidation and the low-hanging fruit workloads have mostly been virtualized. If virtualization only delivered consolidation, we'd be done. But it's moving beyond that use case quickly, as customers leverage virtualization for so much more: application mobility, better data protection and DR, branch office consolidation, desktop efficiency, etc.

The second wave of workload management issues is always harder when a core technology such a virtualization matures. Early virtualization essentially paid for itself in CAPEX savings, but when we virtualize Tier 1 business-critical applications, or user desktops, CAPEX savings take a backseat to application performance and IT efficiency, and this is why we're seeing VM stall around 25-30% overall virtualization.

In 2011 we expect this to be a driving theme. Check out our report on how the virtualized datacenter and the cloud should be designed for visibility moving forward. You can read it here.