All too often administrators see requests for VMs with excessive resources. Application owners worried about performance ask for more than they need. However, virtual machines often don't use all of the resources they're given. Worse, throwing too many resources at a machine can cause it to run less efficiently (for example, there is overhead associated with managing multiple CPUs when only one is actively used). To combat this trend,...
or at least contain it, there are two useful weapons in the system managers' arsenal: showback and chargeback.
As the name implies, showback is used to show the business groups or customers just how much these requests cost the company and how much could be saved from proper sizing. It can be thought of as an educational exercise to show the cost in dollars of what the business asks for.
Chargeback, on the other hand, is the practice of billing customers for what they use. In essence, business groups pay for the resources they use. With chargeback, operational management and reporting tools, such as VMware vCenter Operations Manager (vCOPs), will show IT and business leaders which resources can be reclaimed without affecting service levels.
The beauty of both methods is that rather than being a rough estimate, these calculations are performed using proven formulas and values, designed by experts and calculated by a computer, leaving no margin for dispute.
Showback and chargeback products and concepts have been around for a while. Historically, they have been quite complex to set up and configure. The long-standing joke was that you needed an administrator and accountant in equal measure to make it work. Products such as Digital Fuel, which VMware acquired in 2011, have made it more user-friendly and simpler to install these chargeback systems.
The whole concept is not just about being able to show the amount of resources used, but to show the amount of overprovisioning that occurs. In a very simple example on my home lab, when I ran it as a test, it recovered about a quarter of the entire resources in the test cluster. This doesn't happen immediately, but workload and usage reviewed over time ensure there are no periodic resource spikes.
When this is applied in actual environments, you begin to understand how powerful these tools can be. If your company, like many others, bills for the resources used, you can use the carrot approach and show them the statistics. For example, if over the last six months a business group has only used half of the requested resources, there is an incentive to cut back. If the business group reduced the amount of resources they allocated, it would become less expensive to run the same workloads.
There are several different tools available, and they vary not only in price, but in capability and integration with your existing infrastructure. Vendors' proprietary tools can easily start at five figures and work their way up. The sky is the limit for the price of chargeback tools.
The advice I would offer is: Start off internally and small. Don't try to make a huge splash right away. Make sure that you understand not only how to deploy the product, but also how the accounting works. You don't get a second chance to make a first impression. If the sums are wrong, regaining credibility and deploying a successful showback or chargeback policy in the future will be incredibly difficult. Try it, get a feel for it and once you have your figures, ensure you triple-check everything.
Dig deeper on Server consolidation and improved resource utilization
Stuart Burns asks:
What benefits has your company experienced from using chargeback and/or showback?
0 ResponsesJoin the Discussion