BACKGROUND IMAGE: iSTOCK/GETTY IMAGES
VM provisioning is quickly evolving into a fast and easy task, but this is proving to be a mixed blessing for IT administrators.
Virtualization professionals routinely provision new virtual machines; it's a process that establishes a new virtual machine (VM), assigns privileges to the VM and allocates the computing resources that the new workload will need, such as CPU cycles, memory space and storage volumes. We recently spoke with Dave Sobel, virtualization expert and Microsoft MVP for Virtualization, to get his perspective on emerging trends in VM provisioning.
We opened with several questions about how Sobel sees VM provisioning practices today. For instance, is the proliferation of "cheap" computing resources undermining the benefits of consolidation? Does he see anything else happening with resource provisioning?
Automation is perhaps the biggest trend emerging in VM provisioning: relying on software tools to assess the needs of a workload and recommend (even automatically allocate) the necessary resources to ensure adequate performance with a minimum of manual "trial and error" work from IT staff. The ability to manually adjust VM resources is a critical option, but, Sobel explains, the trend toward automation is good for the industry, simplifying and reducing overhead tasks, and letting data centers scale to accommodate customers' needs even more easily.
Although the proliferation of powerful servers with multi-core processors and extensive memory might foster a lax or wasteful attitude in VM provisioning, Sobel underscores the value of automation as a benefit to consolidation. "Cheap computing resources actually benefit consolidation," he said. "Replacing older hardware is easier to justify, and driving efficiencies in higher density systems is more justifiable."
We asked Sobel for his insights on how an IT administrator can optimize the computing resources assigned to a VM. Is there a tool or process that can help tailor the right amount of resources, or is it really a matter of long-term tracking and analysis, and so on?
There are few definitive guidelines for allocating optimum resources to any particular VM. Consider that every workload has different resource needs and unique usage patterns. The trend toward automation can help take some of the guesswork out of the provisioning process, but automation is far from foolproof.
Sobel notes the actual provisioning is only the start of an ongoing management process that must also include long-term workload performance monitoring and analysis -- often with the same software tool or suite. Administrators can then make careful adjustments to resource allocation to ease resource shortages or reclaim resources that may go unused over the long term.
For example, a workload that consistently uses more than 95% of the allocated CPU cycles may encounter performance problems during periods of unexpectedly high demand, and an administrator can address potential problems by allocating more CPU cycles to forestall bottlenecks. Consistent and proactive management practices are critical to support effective results.
We closed with questions about how Sobel sees trends in VM sprawl. We know the problem, and we talk about reining it in, but we were curious about how he sees the trends and how users are dealing with the issue.
Virtual machine sprawl occurs when VMs are allowed to remain in the data center environment long after the workload is no longer needed. This presents a particular problem, because the workload will continue to use computing resources and impose data protection and backup requirements on the business, even though the workload is not producing any meaningful work. The ideal solution is to remove unneeded workloads, so that their resources can be returned to the resource pool for provisioning to other more important workloads. Unfortunately, many companies lack the staff, policies and discipline to maintain granular control over VMs, and sprawl is the inevitable result.
Still, Sobel sees a bright side to the ongoing problem of VM sprawl. "Sprawl means that users are virtualizing more and more, and workloads are able to be distributed more evenly," he said. "The right management tools will help ensure that IT administrators can address the problem and find VMs when needed." The additional automation that appears to aid provisioning can also bring enhancements to VM lifecycle management, alerting an IT administrator to idle, underused or expired VMs that should be removed.
Provisioning is becoming more automated, allowing IT staff to step back from the daily grind of tweaking workloads and focus on longer-term business projects. But provisioning cannot exist alone; monitoring and reporting features can track VM performance and provide the insights needed to optimize resource allocation. Yet administrators must always guard against wasting resources by preventing VM sprawl through sound tools and policies.