Hello,

There is no solid answer to this type of question. As discussed in the article you linked, it’s a very dynamic situation and the general rule is to “Play nice” with your neighbors. Generally you want to keep it so your monitoring is showing your CPU below 100%. Above these levels and you are using more than your assigned CPU usage and something you want to avoid. This is generally when it starts to cause issues both on the underlying hypervisor and likely within your server as well.

While the occasional spike isn’t generally harmful, continued use will cause problems for everyone. In the end though there is no real solid value because it’s an environment that doesn’t really provide such solid numbers on the upper and lower limits.