Virtualization capacity planning and performance monitoring: two great tastes that taste great together?
Capacity planning and performance monitoring tools are mainstays of the virtualization management market, but as virtual infrastructures evolve, the two are seeing increasing integration, in practice as well as in products.
When it comes to tools, that means capacity planning is moving away from its static, spreadsheet-based roots into the kind of real-time reporting customary in performance monitoring, and performance management is beginning to take capacity planning concerns into account.
When South Carolina-based Horry Telephone Cooperative (HTC) went looking for a virtualization management tool about 18 months ago, it looked for a performance monitoring solution that would provide visibility into storage and network I/O. The company found that Quest Software’s vFoglight satisfied those needs. At that time, capacity planning was not as high on the priority list, said Philip Sellers, a systems administrator for HTC.
However, HTC’s virtual environment has been growing by about 10 to 15% per year, and capacity planning has become part and parcel of performance management, and vice versa. Sellers said HTC is currently conducting a trial on backup tools, for example, where vFoglight’s capacity planning capabilities have become equally important.
“Based on numbers vFoglight is giving us, we’re able to extrapolate how many [virtual machines] we can put per data store,” given expected backup I/O on those data stores, he said. Assuming you know your CPU parameters, “it’s a good tool for forecasting the future.”
The drive to optimize
Industry experts say the union of performance monitoring and capacity planning in users’ minds as well as tools on the market is a growing trend, in part because of the current economy, and in part because of the increasingly dynamic nature of virtual workloads.
“Until recently, customers did not mind over-purchasing their server hardware to make up for not having good capacity planning tools,” said Brandon Worrell, the director of business development for Salt Lake City, Utah-based systems integrator Solutions II. But a recessionary economy means IT pros have to do more with existing assets and require more insights into the environment to squeeze more out of resources.
“We often say, don’t buy that one extra server, buy the capacity planning tool instead,” said Worrell.
Private cloud computing initiatives also contribute to the need for more performance-oriented capacity planning, as well as capacity-aware performance troubleshooting, said Gartner analyst Chris Wolf.
“A lot of clients are doing private cloud initiatives and realize they can’t possibly automate IT as a Service unless they have a good idea what’s there, and tools have to be very dynamic to keep up,” he said. Their goal is to achieve higher rates of virtualization on the same shared infrastructure, including performance-sensitive tier-one applications. “We’ve been seeing a significant increase in pilots in this area,” he said, and “users are making purchasing decisions this year.”
Competition heats up
Vendors including VMware Inc. see this trend as well, and it’s reflected in VMware’s vCenter Operations tool, which can automatically correlate and remediate both capacity and performance bottlenecks under one umbrella. Meanwhile, independent software vendors in the VMware ecosystem are polishing their products to compete -- meaning more choices and potentially better pricing for users.
VKernel, for example, which has its roots in capacity planning, began reporting on storage I/O performance with version 2.2 of its Capacity Management Suite in January. In April, Veeam Software Inc. added a new capacity planning module to its traditionally performance-focused Veeam Reporter software.
Then, earlier this month, VKernel released version 3.5 of its vOperations Suite with more visibility into network and storage performance, including average and peak throughputs, IOPS and latencies. On the networking side, version 3.5 adds visibility into network metrics including usage rates and dropped packets.
VKernel’s unified view into performance and capacity planning was a big selling point for Peer 1 Hosting, a managed hosting provider in Georgia that recently bought the product.
“Over the last two years we have made a very significant investment in our virtual infrastructure,” said Chris DiGanci, an IT manager for Peer 1 Hosting. “While all the tools were there, we didn’t have that one view,” he said.
SolarWinds Inc. also rolled out version 4.0 of its Virtualization Manager product last week, based on the company’s acquisition of Hyper9 Inc. in January. The new release extends capacity planning features to include storage I/O and network I/O metrics.
“We’re finding that as people deploy more mission-critical applications [on virtual platforms], typically storage I/O is the bottleneck that they encounter,” said Jonathan Reeve, the senior director of product management for SolarWinds. Virtualization Manager 4.0 is intended to “help them get ahead of those issues by predicting, ‘How many VMs can I fit in my cluster or my data store from an I/O perspective, [and] when am I going to run out of storage I/O resources?’” Reeve said.
Fetch Technologies, a Web-based business research firm, pushes the limits of server consolidation -- as far as an 80:1 VM-to-host ratio -- with the use of beefy four-socket, 12-core Dell physical hosts and careful monitoring, said Rick Parker, IT director, who purchased Virtualization Manager to understand the correlation between resource capacity and performance capacity. Parker said Virtualization Manager helped him identify the point at which the available physical cores on a host can no longer service demand from all the virtual cores.
“You can increase the number of virtual machines per host till you reach that threshold and then you have to start backing down,” Parker said.
Beth Pariseau is a senior news writer for SearchServerVirtualization.com. Write to her at firstname.lastname@example.org.
Dig deeper on Capacity planning for virtualization