Hi

System activity monitoring tools - top, iotop, ntop, sar, collectl, etc - may be a good reference to judge the system activity when the system transitions to sleep state.
But if I make the system transition to sleep state when i/o activity is zero during 15 minutes, for example, it won't sleep forever because slight i/o by daemons, etc occurs continuously even if no user i/o.
So how can I judge the system activity to change the state by using those tools?

Thanks in advance.

Normally you're not going to be performing detailed system monitoring on a machine that you expect to enter a "sleep state" (which to me implies a laptop or desktop). Can you talk more about the specifics of what you are trying to accomplish?

Thanks for your reply.

I'm going to write a power saving tool which change a machine (laptop or desktop) to sleep state.
The state will be changed automatically when no system activity during some periods, and a problem is how to judge the system activity.
What I'm thinking are
cpu/memory activity
i/o activity
and network activity
,and there are several monitoring tools for each.
So I'm trying to judge the system activity through such a tool, that is, if both i/o read and i/o write are zero during some periods that results from "sar", for example, the power saving tool judge the system is inactive and change it to a sleep state.
But output from "sar" says that both i/o read and i/o write are always not zero. This is because slight i/o by daemons, etc occurs continuously even if no user i/o.

Cpu/memory and network activity are also similar.

I wonder how can I determine the system activity by using such a tool.

This is indeed a tricky one because if your tool is running on the system IT will be generating system activity. Running something like collectl would let you look at a lot of things but it too would generate activity. I'm wondering if looking at the interrupts might help? Of course there are always clock interrupts even on an idle system and other things would also generate noise as well, so this may prove to be an impossible task, or at least a very complex one.
-mark

Good one. To make it even worse not all system activity is equally important that it should keep a machine from entering ACPI sleep state, right?