If left unchecked, virtual machine sprawl will slowly sap your data center resources and kill application perf...
More on VM sprawl
Control VM sprawl in your virtual server infrastructure
Closing the VM sprawl floodgates
Virtualization challenges: Security, storage and VM sprawl
The first step in combatting virtual machine sprawl is recording and tracking the workloads in your environment. It may seem daunting, especially if lax provisioning policies were in place. Armed with this information, you can implement effective policies to curb virtual machine sprawl, once and for all.
Methods for tracking virtual machines
Believe it or not, some admins try to track all their VMs in a spreadsheet. Luckily, there are lifecycle management tools and other strategies to help you avoid this excruciating manual process. That’s not to say a good configuration management database isn’t a valuable resource, but manually tracking and tracing changes via an Excel spreadsheet leaves the door open to poor maintenance and inaccurate or out-of-date information.
At the very least, you should use permissions to prevent rogue administrators from creating VMs at their leisure. You should also create a request form that requires application owners to submit “purchase orders” for new VMs. It sounds draconian and contrary to the “agility agenda,” but you need a system of checks and balances to stop these guys from doing as they please. It’s not enough to simply create a spreadsheet and start inputting data. In my experience, these systems are often poorly maintained. If you also use them to record IP addresses, out-of-date information will inevitably lead to Internet Protocol conflicts.
Of course, you need to keep some record of this information. PowerShell scripts are the best tools for generating weekly reports of an environment. Plenty of free reporting scripts are available, including Al Renouf’s vCheck for VMware vSphere. It generates reports about your environment and flags configuration problems.
If you have the resources, there are commercially available reporting tools such as Veeam Software’s Reporter utility. It has a free edition that can report on VMs by usage, and the tool may help put a dollar value on the potential savings from pruning unneeded VMs.
Also free is Veeam Software’s Business View application. It allows you to categorize VMs by business unit and to see the consumption in a more logical format. Most of these applications support the export of data to an Excel spreadsheet. If you must store an audit trail in .xls files, it’s better if their contents are system-generated rather than manually edited and maintained. You’ll probably also need to refer to a hefty list of physical machines, especially if you want the spreadsheet to contain IP addresses assigned to virtual and physical machines.
Nipping virtual machine sprawl in the bud
The best way to prevent VM sprawl is to develop a sensible deployment policy that won’t drag down performance. Create a modest lifecycle management portal that allows users to make requests for new virtual machines. If your enterprise uses VMware vSphere, VMware vCenter Orchestrator (vCO) may be good candidate.
The vCO system is powerful, and with a license for vSphere 4 Standard or higher, it’s free. The product includes a series of built-in “workflows” and allows customers to automate processes with their own workflows. VMware customers should check out the new vCO Team Blog, which includes videos that demonstrate how to create a simple self-service VM provisioning portal. The second and third parts of that series are available online.
Alternatively, Microsoft Hyper-V comes with a built-in self-service portal that is part of System Center Virtual Machine Manager (SCVMM). The portal comes with controls that the administrator can use to allocate a quota of VMs for each user. This allows you to limit the number of VMs that can be created. As you might expect, Citrix has a self-service portal centered on its own virtualization platform, XenServer. Originally part of the “Cowley” beta project, Citrix’s self-service portal ships as a downloadable virtual appliance of about 200 MB in size and has Linux as its core operating system.
All that Citrix users need to do is download the appliance and add it to an existing XenServer pool. The product uses Microsoft Active Directory credentials to import users, together with their permissions and rights, and it allows them to carry out tasks such as starting, stopping and suspending their VMs.
Note that the tools from VMware, Citrix and Microsoft focus on their own proprietary hypervisors. If you have a heterogeneous environment, you may need third-party self-service portals, such as Quest’s vControl.
Another downside of most free self-service portals offered by the major hypervisor vendors is that they do not include a “chargeback” or “showback” module because they charge for this functionality separately. This is disappointing, because attaching a dollar value to each VM can minimize sprawl. This can help people avoid treating VMs as though they’re free; after all, with virtualization, the creation of a new machine is just a mouse click away. But, in fact, every VM costs money—either in the form of licensing costs for the guest OS and application, or the disk, memory and CPU resources that are allocated.
Managing virtual machine sprawl in the cloud
The next big step from these entry-level self-service portals is a cloud director platform, which gives administrators greater control over how virtual resources are consumed. Most cloud director software comes with chargeback functionality. The goal is to build an IaaS offering, where internal IT can begin to charge for its resources and services just as an external provider would. Over the long term, this is the model most businesses will adopt.
IaaS doesn’t offer the specific services of Software as a Service or Platform as a Service, but it can be a foundation for them. With an IaaS offering, you should be able to offer a so-called service catalog, which is effectively a VM library from which users can select. In addition, the golden master images should form the bedrock of new VMs that end users can customize with additional software.
Ideally, as end consumers define new VMs, they will see the costs they are incurring. In this way, IaaS can really stifle virtual machine sprawl. In the short term, if your goal is to reduce virtual machine sprawl, start with an automated audit trail and reporting system, as well as a product that allows both self-service and an approval engine.
No excuses for virtual machine sprawl
In numerous ways, the shift from physical to virtualized infrastructure has upended traditional management methods. In the absence of virtual server tracking and management tools that really do the job, data center managers have sometimes thrown up their hands and said, “What can we do?” But as virtualized infrastructure becomes more far-flung and complex, knowing the status of your resources is a necessity—not a nicety.
And with new scripts and lifecycle management tools available, there’s no need for data center managers to permit departments to create untracked VMs. Now they can enlist tracking systems that enable them to identify these rogue servers, nip virtual sprawl in the bud and take a more sophisticated approach to managing virtual systems.
Virtualization has massively reduced the time it takes to provision a new server. Now, with a few clicks of a mouse, a new VM can be created. Sadly, that rapid deployment made it too easy to create new VMs without proper authorization and audit trails. The days of hoping that administrators will play nice and obey policies are quickly coming to an end.
The reality is that you need to curb the powers of rogue admins who create new VMs whenever they feel like it. It’s not surprising that virtualization has created VM sprawl because the technology has been very intrusive to existing business processes. Virtualization has also caused many to question whether existing IT divisions—the network team, the storage team and the x86 team—are even relevant. In many respects, we are in a transitional period as the ratio of virtual to physical desktops grows inexorably and we begin to understand the impact of this evolutionary technology.