It’s easy to understand the lure of virtualization and virtual machines (VMs) – new VMs are fast, easy to create, and the hardware is essentially free. If you need to duplicate your
The challenge of virtual machine sprawl
Ease in VM creation can lead to VMs proliferating across systems. When added together, these stray VMs account for a significant amount of overhead for your organization. According to "Understanding virtual sprawl," a whitepaper from Embotics, "an environment of 150 VMs will have anywhere from $50,000 to $150,000 locked up in redundant VMs."
Having stray VMs is unavoidable. But as more are added, administrators shift their attention to the management and upkeep of these individual systems and away from the overall health of the environment, and are less likely to notice sprawl becoming bigger.
Unfortunately, this becomes a vicious cycle. Once a critical mass of managed systems is reached, administrators are likely to reprioritize their management workloads and only manage what's needed. They move on to active VMs and let older, unused VMs go unmanaged. While this alleviates their day-to-day systems management workload, the old VMs still take up physical and financial resources. Unused VMs, even if powered off, may still have software licensing and support costs tied to them. The cost of unproductive computing resources and software licenses can become an organizational burden.
And it’s easy to exacerbate this waste. Administrators often overprovision VMs, providing a fixed amount of memory, storage and processor cores regardless of a VM’s needs or purpose. In addition, snapshot backups often accumulate over time and take up extra storage to protect an unneeded VM.
Reining in virtual machine sprawl
So why is it hard to decommission old VMs and reclaim physical and fiscal resources they use? VMs are easy to manually create, but must be manually removed. That manual process competes with the day-to-day requirements of managing the rest of the network and can get pushed aside in favor of immediate needs. This is true even when using products like VMware vCenter to manage VMs.
Sending out a team to actively decommission VMs can also be problematic, because VMs can be difficult to find depending on state. Finding the owner of a given VM, for example, and figuring out if the VM is still needed can be a time-consuming and error-prone process. Even if these teams manage to find and decommission VMs and reduce or remove virtual machine sprawl in an environment, it's important to implement best practices to prevent the problem from reoccurring, which include forming a specific plan for building, maintaining and retiring any VM created in your environment. If VM retirement becomes routine in VM creation, then the odds of sprawl are decreased. But don't just think about VM retirement -- controlling the VM creation rate is also excellent for slowing or halting sprawl. Create policies for building VMs -- not just retiring them. Have solid business policies in place to justify creating any VM in your environment. If managing VMs manually, be sure all VM owners are aware of these processes and are fully engaged in participation.
Most VM management solutions point to a lack of automation as the root cause of virtual machine sprawl. VMs are often created manually based on a template or customized for a specific need. If VM discovery and tracking across a network is automated, then overhead management is reduced and managing VMs becomes much easier.
Embotics' V-Commander tool, for instance, provides an automated way to locate, track and deliver business, technical and fiscal information about any VM within an environment. Lifecycle management is a key part of V-Commander, which provides a great high-level view of VM management.
By comparison, VMware's vCenter Lifecycle Manager (LCM) application is undergoing significant changes, which raises questions about how vendors plan to manage virtual machine sprawl going forward. According to VMware, LCM is no longer being sold separately as of Sept. 1, 2010, and support for LCM will formally end March 1, 2012.
VMware is folding the functionality of LCM into cloud management products like vCloud Director and the addition of LCM will add governance and provisioning features to vCloud Director. In the vCloud framework, virtual resources are managed more as services and not as individual VMs. Services are deployed as needed and are based on administrative policies, using just the computing resources required. Once the services are no longer needed, they can be retired, as can the VMs.
The Infrastructure-as-a-Service model for VM management offered by vCloud Request Manager, due out by year-end, may also help manage VM sprawl, particularly for larger organizations. Smaller organizations should be able to use a server-level automated solution, such as V-Commander, to keep virtual machine sprawl in check.
Managing virtual machine sprawl takes business discipline
It’s clear that avoiding virtual machine sprawl requires regular maintenance, rigid tracking or an automated tool to manage the problem. As VMs are so easy to create, thinking about VM sprawl before it becomes a problem is also important. Implement comprehensive creation policies to help stem the flow. With potentially hundreds of thousands of dollars at stake, ignoring VM sprawl is definitely not an option.
What did you think of this feature? Write to SearchDataCenter.com's Matt Stansberry about your data center concerns at firstname.lastname@example.org.
This was first published in November 2010