One of the biggest problems with 1st party cloud portals is their inability to effectively communicate unused cloud resources, and their associated monthly costs.
This the equivalent of not realizing that you may have left lights on in another room. These resources constantly raise the overall overhead cost of running your service.
This week, we're releasing our Idle VM detection and notification feature. This service notifies you whenever VMPower detects a virtual machine which has been idle for 1, 3, 10 or 30 days, depending on your preference.
Detecting Idle VMs
So how do we do this?
VMPower constantly monitors your VM instance statistics by tracking data points such as cpu percentage time, memory and disk read/write speeds. For AWS this is implemented via Cloud Watch free tier usage. On Azure, this is backed by diagnostics data that Azure agents log to storage tables.
We make this information internally searchable by the VMPower analyzer. This allows us in the future to release fast predictive analytics, metrics and report high level insights on basic patterns of cloud utilization relative to overall capacity.
In the case of an idle vm, if the vast majority (approximately 95%) of data points fall under the threshold mark, VMPower identifies that instance as idle. Afterwards it sends a notification to you and displays this on the platform:
The graph above shows a typical idle Windows VM from Azure overlaid with a blue line at the 2% CPU time mark. Over a 12 hour period, this VM had only 2 points where the VM was recorded to have a CPU time that exceeded 2% (denoted by the blue line).
When VMPower identifies instances that exhibit this behavior, they will appear on your dashboard and you can optionally browse the VM performance chart by clicking the chart button:
Based on this information you may either deallocate or place the VM on an automated allocation schedule.
Once you enter your cloud subscription details to VMPower, you will begin to identify idle VMs within 24 hours.