Along with giving enterprises great ease and speed of provisioning their IT capacity, virtualization technology has brought along with it an entirely new set of problems. Specifically, enterprises are finding themselves with huge amounts of virtual machines that are extremely difficult to manage efficiently.

Shawn Edmondson, Director of Product Management at rPath – provider of application deployment and maintenance automation solutions – proposed his company’s suggestions for managing what he called “VM sprawl.”.

“Sprawl is when (the number of) VM instances are growing faster than you expected or faster than you wanted,” Edmondson said. “All you have to do is fire up VMware’s client tool and you do a few clicks and you pick a template and – 15 minutes later – you have a new VM up and running in a cluster that you can connect to and run your app on. You end up creating lots of new ones without actually shutting them down or reusing machines that are actually out there.”

What causes VM sprawl?
Ease of provisioning, however, is not the only reason for VM sprawl. Other forces are at play as well.

One of these forces is increased “software diversity.” Enterprises no longer deploy applications that are monolithic, but very complex applications that have countless numbers of various components. The rate at which organizations crank those applications out has also reached incredible speeds. Plenty of enterprises out there release one or more new applications every day, with many releasing up to 500 applications every year.

Shawn Edmondson: "Sprawlers are not trying to break the infrastructure. They just want to deploy their applications and they just want to achieve their business outcomes."

The final reason for VM sprawl is the sheer scale today’s enterprises operate at. “Going from single computers to departmental servers to whole-enterprise servers to … the cloud, it’s now easy to have an application that’s running on hundreds of servers, spread out across the entire world.”

“It’s a collision of complexity,” Edmondson said. “Fundamentally, sprawl is a natural consequence of all this complexity colliding with all this demand for applications from a business.”

Taming the sprawl
rPath’s proposition for dealing with the sprawl has three basic components: reducing per-instance cost of managing VM’s, greater standardization and application-centric IT infrastructure design (as opposed to OS-centric).

Per-instance cost can be reduced through smarter automation. Virtualization and cloud computing have placed more demands on automation solutions than there were in the past. Today, automation needs to be able to handle VM images and short-lived, as well as long-lived systems, in highly dynamic environments.

“With good automation you can improve 10-fold the number of servers that one system administrator can take care of.”

The goal of greater standardization is to have fewer unique VM configurations. “If you have 5,000 VM’s, it sounds very frightening,” Edmondson explained. “But if they’re all the same, it’s really not that bad.”

Of course, it is impossible to apply one image template to all VM’s deployed. Smart standardization entails finding “a flexible way to combine standardized bits and pieces and get the best of both worlds. You need the flexibility of machines that are different with the standardization of having machines that are the same.”

Application-centric IT infrastructure design is the third basic principle in navigating VM sprawl rPath suggests.

“Before virtualization, before the cloud, operating system was like an apartment building and the applications were like tenants,” Edmondson said. “The servers and the operating systems were persistent: they ran for a long-long time; they went through many OS upgrades and patches and configuration changes over time; different applications were installed or removed over time.”

Virtualization and cloud computing have removed the necessity to do that. Instead, IT departments need to focus on the applications themselves, letting the OS fade into the background. “It can be more like a supporting library that’s part of the application, instead of this dominant thing that we focus on.”

Related analysis: Reconciling Virtualization and Consolidation with Availability and Power Efficiency
Related white paper: CA: Virtualization and Automation Drive Dynamic Data Centers
Related news: Cisco discusses its game plan within the cloud