Page 2 of 3   « previous page   next page »

Is Sprawl More Likely in the Virtual World?
Sprawl is not defined by the number of virtual machines. Many organizations have thousands of VMs that are not in sprawl condition. Some organizations have less than a hundred VMs and are in a state of sprawl. The definition of sprawl, “ the proliferation of virtual machines without adequate IT control,” has more to do with control than numbers. If growth is not controlled, then the organization is in danger of sprawl.

At a recent speech at Interop in New York, about 5% of approximately 90 IT professionals in the audience knew how many VMs were running in their organization. If they cannot be seen, how can they be managed or controlled?

Frequently the management and tracking of VMs is done manually. During the same Interop discussion, when asked what was being used to manage and track VMs the number one answer was “spreadsheets.

VMs break the traditional processes and controls that data centers have to control physical server deployments. These processes typically involve a number of different teams all signing off before a server is ordered, provisioned, and deployed. The process ensured that no deployment occurred if the various signoffs were not obtained.

Conceptually, if an administrator can see a server on the network, a VM can be deployed to it – with no process-driven checks and balances.

There are also inevitably more VMs in any given environment than in a traditional “physical” data center. VMs are not only easy to deploy, they are almost universally considered “free.” Once the host is paid, the cost does not change if you have five or 25 VMs on it. Also, once you have gone through the initial server consolidation project, there is lots of unused capacity available anyway.

Business owners have not only woken up to the fact that servers can now be “instantly” deployed, but they are also “free.” Subsequently, they are almost universally doubling up to provide redundancy and the assurance of no unplanned downtime. New VMs are routinely deployed to cover demand spikes, or to maintain a level of service during maintenance or troubleshooting.

This combination of poor visibility, manual administration, lack of deployment process, and the comparatively large volume of VMs all work together to make virtual sprawl more likely than server sprawl, and in some cases almost inevitable. In no time, hundreds or thousands of virtual machines are running on the network, consuming memory, disk space, and CPU cycles, half of which are lost track of or stand unknown why they were created in the first place.

This is especially true where multiple groups are creating VMs without centralized control.

Why Should I Worry About Virtual Sprawl? Does It Really Cost Me Anything?
Anyone who experiences server sprawl knows how expensive this can be. Virtual sprawl is no different.

As with server sprawl, the costs start with wasted physical resources, including direct resources like systems, disk and memory, as well as indirect resources like power, air conditioning, and rack space.

Add to this the direct cost of wasted software resources, including application(s), virtualization platform and operating system licensing, as well as the cost of any management agents that are deployed inside the VM. There also may be some indirect licensing costs that apply here. For example, tools such as OpenView can have licensing models that are priced by node, independent of agents. The more nodes, the higher the cost.

Sprawl also creates increased administration costs due to the need to manually manage the environment. An example of this is dealing with the patching and updating for an environment that is larger than needed or manually producing reports for auditors or upper management.

Having a larger environment than is necessary also creates increased risk for an organization. Growing VM environments without adequate control provides a greater potential for unauthorized or “rogue” VMs to creep into an environment unnoticed, or for unsupported OSs and applications to proliferate, resulting in the potential for network, configuration, interoperability, performance, capacity, or service-level conflicts.

With a wider variety of virtual machines in the environment, when there is a need to re-create a virtual machine, there is a greater likelihood of creating one that does not comply with corporate standards or regulations. This has costly license payment implications or one that is subtly (or not subtly) different from the one created for the same purpose last time.

For the most part, corporate IT auditors have not woken up to some of the differences between physical servers and virtual ones. But when they begin to, the need not only to be in control at the time of the audit, but also to demonstrate control over the whole audit period is essential.

Page 2 of 3   « previous page   next page »