Cluster efficiency is available as part of the cluster report. It provides helpful information on CPU and MEM usage in your cluster, letting you assess its efficiency and plan action to reduce overprovisioning. Read on to learn how to use it.
The report delivers insights on your cluster’s CPU and MEM usage and helps you put a number on overprovisioning. You can browse current and past data:
If you select the current period of time, you get real-time insights on your overprovisioned resources.
You can also see exact numbers on your cluster’s provisioned and requested resources, as well as the average hourly cost per CPU and GiB. These numbers don’t change if you select dates in the past.
Then, you can delve into details on your overprovisioned resources and correlate that data with costs, including provisioned and requested CPU/GiB and their difference.
Using the lifecycle filter, you can see how these results differ across spot, on-demand, and fallback instances.
There is also an option to see hourly, daily, and monthly cost rates of the selected resource.
Finally, depending on the type of resource you select in the section above, you can delve into daily CPU and memory efficiency details:
Each day provides calculations of cost per provisioned and requested resource, together with the exact overprovisioning rate reached on that date.
By clicking the arrows by the name of each column, you can shuffle the results to see the days with the highest/lowest overprovisioning rates or costs per provisioned and requested resources.
Overprovisioning – this rate refers to how many provisioned resources you could reduce. It follows this formula: 100% - (req/prov * 100%)
For example, if you provisioned 688 CPUs and requested only 490 CPUs, then overprovisioning would reach a rate of 28.78%.
Updated about 3 hours ago