Hi All,
After an attempted upgrade of our production BIG-IPs (i4600 HA pair) from v14 to v16 earlier this year that required a rollback due to high CPU utilization, I have been doing testing on some VEs and noticed that v16 has about 15% higher CPU utilization just from monitors.
The Setup: only LTM and GTM are provisioned, but no GTM usage. I copied config from production units to the test VEs, excluding VIP and SNAT objects. The LTM has around 1200 pools configured, 2-4 members in each pool, 1-2 monitors on each pool with most only having 1 monitor. The majority of the montitors are HTTP/HTTPS, which I understand do use more CPU than something like TCP (https://my.f5.com/manage/s/article/K42421592).
However, it seems odd that monitors on v16 would use almost 15% more CPU than prior versions. I was running v14 as a baseline, noticed nominally higher (~1-3%) CPU in v15, and around 15% in v16. Has anyone else noticed this, or able to provide an explanation?
Below are some screenshots of the CPU graph on the VE when switching betwen v14.1.4.6, 15.1.9.1, and 16.1.3.3. The first is a 24-hour window, and the second two are 4-hour window graphs.
