Cloud servers are easy to provision and configure. Maybe too easy. That’s why many organizations are finding their cloud spend spiraling out of control. If you have recently experienced shock and awe at your monthly cloud bill, you may need to examine your environment for optimization opportunities.
Here are four of the top areas to reduce your cloud sprawl, and by extension, your cloud spend.
While it may be simplest to set up your cloud environment at the maximum resource levels you’ll need for peak usage, this is the fastest way to rack up unnecessary charges. Part of the appeal of the cloud is being able to add additional resources on demand, so take advantage of resource monitoring and capacity planning to rightsize your VMs. This may take a few cycles of experimentation, but eventually you’ll hit a sweet spot that provides the performance you need even for minor peaks while still avoiding overprovisioning.
If you start to experience a resource crunch and VMs are failing, vCloud does include Hot Add features to provision additional CPU or memory for individual or grouped VMs. There are limitations to this however, so be sure to read about what is and isn’t possible when using Hot Add. If your cloud is not VMware-based, you may be able to set up autoprovisioning features.
Most every VMware user eventually runs into the problem of orphaned VMDK files, or unused virtual machine storage disks that are consuming space — and therefore cloud budget — while not being actively used by a VM. VMDKs do not have any indication of which, if any, VMs are using it, so once your environment reaches a certain scale they can be difficult to locate.
Old snapshots can also consume valuable storage space. If you’ve snapshotted a VM more recently than an older snapshot, chances are you don’t need to save it. There are exceptions in the case of requiring multiple versions of backups, but eventually you’ll reach a state of snapshot clutter that isn’t sustainable for your budget.
There are freeware tools like RVTools that can help you find orphaned VMDK files and old snapshots. Be sure to also train your administrators to remove virtual disks when old VMs are decommissioned.
Speak with an infrastructure consultant today.
If you’re using the cloud for software testing and development, or you have test VMs spun up for any reasons, like demoing or testing a new application or a new version before rolling it out to production, you are likely only using those VMs during work hours. While test VMs are convenient, they can also easily lead to cloud sprawl. If you leave a test VM running for a extra week and you only needed it for the work week, that’s 28% of your infrastructure expenses wasted for that instance.
This speaks more to your test/dev process and operations than anything else. Be sure to carefully track your test VMs and shut them down when they are no longer needed. If you think you might need a VM again, you can always pause it or move a snapshot to a storage archive.
You may have a number of virtual machines that are powered on, but have idle or underused workloads. In this case you are overpaying for resources that are going unused. A good rule of thumb is to run VMs at around 80% utilization, so they have some room for spikes in demand but are generally an efficient use of memory and processing power. It is key to have a proactive monitoring strategy for this to be effective, as you don’t want machines to crash under sudden demand.
vCenter Operations Manager can help you find idle and underutilized VMs with a default report. It can even run the report and send you results on a daily, weekly, or monthly basis. You set the thresholds for idle CPU, disk, and network usage over a given reporting period. So you may want to identify certain VMs that used less than 100 Mhz in the past two weeks, for example.
These are just four of the most common reasons users suddenly find themselves with massive cloud expenses and underutilized environments. To cut down on cloud sprawl and monthly charges, be sure to regularly monitor and clean your virtual data center of unused or underused resources, and to plan and manage your virtual machines carefully. A periodic audit can be a good way to stay on top of this.