To end users, a private cloud is meant to seem like a bottomless pool of resources. But for the IT professionals who manage that cloud, careful capacity planning is necessary to keep those resources available.
The good news is that effective private cloud capacity planning is within reach, provided you use the right tools for the job.
"Your Excel spreadsheets are not going to cut it anymore," said Chris Wolf, research vice president at Gartner Inc. of Stamford, Conn. "It's funny, because a lot of clients still look at it that way, and that old model of planning capacity by just how much processing and memory and storage you have doesn't fly when workloads are constantly moving from one physical server to another."
Experts recommend a variety of tools for capacity planning, from BMC Software's Capacity Management and Capacity Optimization tools to VMware's vCenter Operations Management and CiRBA's Capacity Control Console. Private cloud capacity planning tools are also available from Microsoft, IBM and EMC. Open source tools are in use as well, including Graphite and Ganglia.
Any of these tools would be an improvement over manual methods, said James Staten, vice president and principal analyst at Forrester Research in Cambridge, Mass. "At this point, the traditional capacity-planning tools are probably the best, and we see them more in use than anything else," Staten said.
Chasing the zombie (VM)
Reclaiming capacity from inactive machines becomes especially important at scale, said Brian Cann, vice president of cloud and infrastructure for CGI Group, an IT and business process services firm based in Montreal. In addition to the large public cloud CGI offers its clients, the company has a private cloud environment consisting of thousands of virtual machines. Cann said CiRBA's tool helps him curtail "zombie VMs" in that private cloud.
"We build up virtual environments and people move off onto other things and leave the environment hanging around," he said. "With the CiRBA tool, we get continuous reporting on those environments and this is a way we can go back and say, 'There's no usage on these environments.'"
Maintaining control over capacity use is also a priority for Greg Ericson, chief innovation officer at Press Ganey Associates, a health care services firm based in South Bend, Ind. For Press Ganey, the most effective approach to private cloud capacity planning is a combination of VMware's vCenter Operations Manager tool for ongoing monitoring of the infrastructure and a governance process to regulate business units' use of allotted capacity in the private cloud.
"We try to handle some of that governance up front," Ericson said. "We have an architecture team … representing the specific application as well as the infrastructure, that goes through and reviews the plans that the teams have."
From there, vCenter Operations will alert the team if any oversized VMs are provisioned.
Gartner's Wolf also emphasized the importance of good governance in capacity planning.
"Just because I have capacity in one cluster for my private cloud doesn't mean I can actually use it," Wolf said. "There might be a security constraint or compliance constraint that prevents a workload from running on a certain physical server, and these nontechnical factors are really important. They're something that capacity-planning software needs to be able to keep up with."
It's OK to think small
Another important factor in private cloud capacity planning is to right-size the environment and avoid overprovisioning, Forrester's Staten said.
"Most companies have a tendency to have traditional capacity planning thinking, where you build as big a cloud as you would ever need from a total size perspective. And you don't actually want to do that with a private cloud," he said. "You want to build as small a cloud as you possibly can, so you can maximize the use of resources."
Network latency and throughput should also not be forgotten in the process of planning capacity for a private cloud, although many IT shops use separate tools for monitoring the network from those for monitoring compute capacity. Press Ganey's IT team, for example, uses SolarWinds' Orion tool suite for network monitoring alongside vCenter Operations Manager.
In general, philosophies around private cloud capacity planning are changing, said Jai Malkani, director of cloud management products at VMware.
The biggest mistake organizations make is treating capacity as a separate discipline from operations, having one team focused on operations and another looking at capacity planning, Malkani said. "Eventually … you're not going to be able to control how much capacity gets provisioned. End users will deploy applications and deploy VMs, which may or may not be out of your control. … From a technology perspective, adopting tools [that] are based on analytics and can deal with the dynamic nature of the environment is very critical," Malkani said.
Dig Deeper on Capacity planning for virtualization