Back in the early days of virtualization adoption, VM sprawl and over-provisioning were quite common. It was simple to create new VMs by effectively copying and pasting. VMs were also frequently over-provisioned, because they were deployed with the same resources as physical servers.
We still regularly encounter VM sprawl in enterprise data centers. But performance optimization and rightsizing solutions allow data center admins to identify and remediate these situations. VM sprawl prevention has become possible through workload lifecycle automation and making up-front costs visible to IT service consumers. No data center is perfect, but the combination of third-party solutions and maturing IT virtualization processes has certainly made a huge dent in on-prem sprawl.
Then along came the public cloud. And to make matters worse, very often an organization’s initial public cloud consumption isn’t even managed by IT. Instead, it’s controlled by various lines of business (LOBs) through shadow IT—teams that have little or no expertise in IT processes. It’s no wonder some organizations experience sticker shock when they get their first AWS or Azure bill.
Some of the solutions for your on-prem data center can apply to the public cloud too. But emerging third-party solutions provide performance rightsizing for various public cloud vendors, allowing you to ensure that you’re purchasing correctly sized instances. Some automation solutions also provide the same costing and lifecycle capabilities for both private and public cloud environments, ensuring a proper rental model for workload consumption. In fact the public cloud is actually driving private cloud costing, as I’ve previously written here.
But there are enhancements that help control public cloud VM sprawl specifically, notably power scheduling. Many of the public cloud workloads are used in dev/test, with engineers leaving them powered on when not in use. A simple schedule that powers these off during non-business hours could potentially save over half of your monthly bill.
I also believe the eventual goal is for IT to manage all private and public workloads and to become the broker for providing ITaaS to various clouds. That said, in many cases the horse is out of the barn with shadow IT, and achieving this brokering goal may take some time, or be nearly impossible in some organizations. In the case where public cloud consumption is controlled outside IT, the IT team still requires a single pane-of-glass solution to, at a minimum, inventory and report on those workloads. Some automation solutions also provide the ability to assign organizational ownership to new workloads as they get spun up. This will guarantee that there is at least some accountability to the LOBs.
Providing ITaaS across all private and public cloud consumption is the goal for most organizations. Along the way, one key stepping stone is properly controlling your public cloud VM sprawl.