While the cloud eases many administrative tasks, it doesn’t absolve IT from basic management responsibilities. And capacity planning is one of those.
CIOs may erroneously think that they can scale back this process since they no longer have to know, down to the terabyte, how much storage and processing power they’ll need to allocate for future growth. They assume that the cloud service provider, using the almost ubiquitous idea of cloud bursting, will allocate appropriate resources. Not so.
While this sounds great in theory, CIOs can’t assume it will happen. In fact, they must continue to own the process of capacity management — especially in a hybrid cloud environment — integrating it with tasks such as provisioning, workload placement and migration, and performance monitoring. The goal is to strike a delicate balance between maximum infrastructure utilization and optimal service delivery to consistently exceed Service Level Agreements (SLAs) and drive business value. If they don’t achieve this balance, say experts, departmental users will simply purchase and allocate cloud-based services on their own. After all, if IT doesn’t allocate enough bandwidth, internal and external storage, and application seats, performance will suffer, affecting productivity and end-user satisfaction.
While you may think about architecture differently in a cloud environment, “You’re still going to need test environments, and you’ll need to plan for the growth of software and the hardware that supports it,” Lora Cecere, Partner and Analyst at The Altimeter Group research firm, told me. “Business analytics, performance and memory analytics, and virtualization software [need to be kept] in-house,” she adds, and they require a specific type of capacity planning.
Automation’s Role
Virtualization may make capacity planning even more crucial since storage growth can “hide” in virtual data centers. Henry Steinhauer, a Capacity Planner at Presbyterian Healthcare Services based in Albuquerque, N.M., says storage virtualization is something his company struggles with. His virtual machines (VMs) request storage automatically when they reach a certain, pre-set threshold, but that can cause problems, he says. “Rather than having the applications specifically request more storage, it’s being requested through VMware, so the cause for storage growth is somewhat masked,” he says. “[Storage] automation has not quite caught up with the [application] request process,” but it is critical.
Just as in the past, IT has to examine and plan for service performance assurance, making sure provider SLAs are met. Cloud computing relies on statistical multiplexing, load-balancing mechanisms, and multiplexing algorithms, so the CIO must focus on these functions and the impact they’re having on the overall environment. Automation is the only way to handle these functions, since agile and virtual applications may move from system to system, changing the way the data is interpreted by the applications.
Steinhauer says he uses automated capacity planning tools that help him create a baseline “normal” for a specific VM or cloud so he can watch for problems or lags with applications. “We’re thinking, ‘Do we expect this to be happening, or is it outside the norm?’ We know that different applications have different profiles, so we can pinpoint potential capacity problems ahead of time,” he says.
Tony Lock, Program Director at the Hampshire, U.K.-based research firm Freeform Dynamics, says Presbyterian Healthcare Services is among the minority; few CIOs have deployed automation in either the active capacity-management process or VM deployment. Most are using capacity management to ensure that their organizations have sufficient resources to meet relatively predictable on-site requirements — not for the dynamic nature of virtualization. This will have to change, Lock says. “Internal cloud models potentially offer great flexibility, but they do not provide infinite resources.” There will be times “when resource capacity will be in short supply and decisions will have to be made on which workloads could be moved onto an external cloud or shut down,” says Lock.
What are you doing to ensure that you’re planning correctly for the cloud? Have you automated these tasks? Share your experiences and successes.
Karen Bannan is executive editor of Smart Enterprise magazine