You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Is your feature request related to a problem? Please describe.
In recent years I have heard more and more that Load Average is a poor metric, often in articles introducing the linux kernel feature PSI.
Describe the solution you'd like
Integrating PSI to btop would be good. Perhaps it could go in the same spot where load avg lives at the moment, although it would be multiple lines ("some" and "all" for cpu, io, and memory; with an avg over 10, 60, and 300 seconds). It's also available for every individual cgroup, but btop doesn't really have anything for cgroups so I don't think it matters yet.
It would be nice to be able to graph these values as well, but with such a big selection of values it's difficult to make a case to make the top or bottom cpu graph portion one of the values as a graph.
Describe alternatives you've considered
Watching the files on the cgroup filesystem manually is "an extra thing" to do, whereas btop is a nice "all in one" solution.
Additional context
Here's a big description mail belonging to version 2 of the psi patch, including "real world application" and the FAQ "what's the difference between load average and cpu pressure":
Is your feature request related to a problem? Please describe.
In recent years I have heard more and more that Load Average is a poor metric, often in articles introducing the linux kernel feature PSI.
Describe the solution you'd like
Integrating PSI to btop would be good. Perhaps it could go in the same spot where load avg lives at the moment, although it would be multiple lines ("some" and "all" for cpu, io, and memory; with an avg over 10, 60, and 300 seconds). It's also available for every individual cgroup, but btop doesn't really have anything for cgroups so I don't think it matters yet.
It would be nice to be able to graph these values as well, but with such a big selection of values it's difficult to make a case to make the top or bottom cpu graph portion one of the values as a graph.
Describe alternatives you've considered
Watching the files on the cgroup filesystem manually is "an extra thing" to do, whereas btop is a nice "all in one" solution.
Additional context
Here's a big description mail belonging to version 2 of the psi patch, including "real world application" and the FAQ "what's the difference between load average and cpu pressure":
https://lwn.net/Articles/759658/
The text was updated successfully, but these errors were encountered: