Increasing demands on busy virtual hosts often call for purchases of new, virtualization-optimized hardware. Chargeback systems, which provide utilization data to ensure fair billing for compute resources, then become particularly useful in justifying and allocating those costs.
A good chargeback approach can also keep departments from overspending. “Implementing a chargeback system forces virtual machine owners to re-evaluate their resource needs to see if there is any way to reduce the amount that they are using,” Eric Siebert, senior systems administrator for restaurant chain Boston Market Corporation, told me recently.
Alex Barrett and I recently talked with Alex Bakman, president and CEO of VKernel. Barrett wrote about a 2007 Bakman interview in the case for chargeback and virtual appliances
VKernel’s chargeback model
VKernel’s chargeback model is borrowed from lessons Bakman learned while working with mainframe chargeback – chargeback based on actual resource consumption. As the developer of CleverSoft and Ecora Software, Bakman had two successful software ventures under his belt before launching VKernel.
VKernel’s Chargeback Appliance is currently in version 1.3.3. It operates like this:
- The administrator defines the various consumer groups as folders, and drags and drops the virtual machines that they use into their folder.
- Costs are calculated either manually or by a spreadsheet available from VKernel’s website. The form asks the administrator about prior purchase costs and suggests the amount the IT department should charge for memory and storage use.
- Finally, the appliance monitors each VM’s activity, and reports usage on a weekly, monthly, or quarterly basis, in a similar fashion and appearance to an electric bill.
The 2.0 version, slated for release later this year, will include the Fixed Plus Overage chargeback method as well as new report types, Christian Simko, director of marketing communications for VKernel told me. The Fixed Plus Overage method enables system administrators to establish a fixed cost for virtual machine use via the current method and bill for overage costs as necessary.
VKernel’s products are modeled in a way so that the system administrator can pick and choose the systems management products they want and need to use, rather than purchasing an entire suite and having to spend time learning how to use it in order to get one or two functions out of the product’s capability.
Ease of use was Bakman’s goal, and the appliance model made that possible. Other chargeback systems he’d reviewed produced complex resource usage graphs which required many hours and a lot of expertise to analyze.
“At one university,” Bakman told me, “an IT department hired an intern to analyze the resource consumption to optimize performance. He quit after 20 days.” In general, he said, systems administrators are “overworked, overburdened and don’t have the time.”
IT pro Rick Vanover chose the VKernel appliance for its ease of use. “As a system administrator, I don’t have time to be an expert in everything,” said Vanover, system administrator for a worldwide vehicle glass repair and replacement company. “The VKernel appliance is easy to get up and running quickly, and it delivers usable information.”
A peek ahead
VKernel also sells a Capacity Bottleneck Analyzer virtual appliance and has two more coming down the pipe. The first, named Search, is slated for release in late August and will allow admins to search their virtual infrastructure for returns on a variety of queries, including, for example, which virtual machines have VMware tools installed and which don’t.
Capacity Modeler, which will debut at VMworld 2008, will allow admins to model changes to the virtual infrastructure before implementation and the outcomes of those changes on the virtual infrastructure as a whole.
Pricing on both Capacity Bottleneck Analyzer and Chargeback is as follows: $299 per CPU for the enterprise edition, $199 per CPU for standard, with support plans at 20% of the total purchase price for one year, or 15% per year for three years if paid in advance. The average total purchase price ranges from $5,000 to $10,000 according to Bakman.