You're almost certainly seeing the effects of launchd starting the periodic daily, periodic weekly, and periodic monthly routine system maintenance jobs, which do things like roll the logs and delete unused temporary files. Those jobs usually run at 3 in the morning, but if your machine was asleep at that time, launchd starts them the next time your machine wakes up.
The job that probably runs the longest and slows the system the most with a lot of disk I/O is the weekly (3AM Saturday morning) makewhatis job, which updates the filename database used by the traditional Unix locate command (sorta like an old-school mds/Spotlight). You can disable this in /etc/defaults/periodic.conf by changing the line
weekly_whatis_enable="YES"
to
weekly_whatis_enable="NO"
If you've added or modified a large number of files on your system, the next time makewhatis runs, it may take a long time. If you kill it every time and it never completes, then the next week it'll have to start over again. I'm not usually bothered by makewhatis, even though my system is not fast, especially my disk. The only time I really notice it is when I've added a huge number of files to my system, like after checking out the entire source tree of the NetBSD operating system.
If makewhatis and the other periodic jobs don't turn out to be your whole problem, be aware that launchd provides the defer-to-the-next-wake service to any launchd job that uses the StartCalendarInterval key. Here's a way to quickly see which launchd jobs on your system use that key:
grep -r StartCalendarInterval {/System,,~}/Library/Launch{Daemons,Agents}/
(NB: I think the curly braces in that statement may be a bash-ism)
