LinkedIn
Twitter
Hello everyone,
I am experiencing CPU peaks during disk demanding tasks on the GCP Compute Engine every 10 minutes. I want to understand the reason why these peaks occur. My goal is to either eliminate these peaks or ensure that they do not potentially affect my application's performance.
I conducted two tests on the GCP's e2-standard-2 Compute Engine with SSD and DigitalOcean's Basic Regular 4GB 2-core VM with SSD for comparison. Both machines run on Ubuntu 22.04.
The tests lasted for 1.5 hours (1 hour with disk load and 30 minutes idle). I used the same bash script on both machines, utilizing fio for disk load, sar for collecting metrics, and gnuplot for drawing the plot. Here is the link to the script: cpu-disk-load-test.sh
On the GCP plot above, we can observe system-level peaks occurring every 10 minutes (yes, there are some additional peaks in the image, but the main repeating pattern, which I derived from multiple tests, is the 10-minute pattern). There is also one peak after the 11:10, even when there was absolutely no load from my side.
Here is the plot from Digitalocean VM running the same script without these peaks:
Although the CPU utilization in general looks different on GCP and DO, due to the different hardware or some other reasons, my main concern here is about these peaks and not about performance.
If you have any ideas why this could be happening, I would appreciate any help.
Thanks!
