On Thursday 06 December 2007, Garrison Hoffman wrote: > Chris Knadle wrote: > > hellanzb is an automated Python Usenet news reader; automated decoding > > of Usenet news can definitely be a process that can generate load. > > Sure, but as I said the Load I am interested in is idle, hellanzb had > nothing in it's download or processing queues.
You did say idle, but it wasn't clear if you were only talking about MythTV being idle. Wasn't clear why killing "idle" processes would drop the system load, even though I know that's possible. > > The other two programs sound like they're related to an IR remote, so > > I don't know how those could be culprits. > > I suspect irxevent, which allows the remote to control the mouse cursor, > the other is just for launching arbitrary programs. That's a little more interesting. You probably had contention there, because two programs were trying to listen to the same IR sensor for input. Even though these programs were "idle", they still could have both been polling for input, so even at idle they could actually have been very active. Much better if the input was gotten via IRQ rather than polling, but then there's possible IRQ contention. > > Try gleaning the LKML [Linux Kernel Mailing List] concerning the > > recent change in scheduler to the CFS ["Completely Fair Scheduler", which > > frankly is an oxymoron]. That was the umbrella subject of threads where > > developers posted output statistics about memory page blocking/locking > > and various wait states as they tested the new scheduler for 2.6.23. > > O'reilly Digital Media has a series on Myth where they liked the > deadline scheduler but reported no great performance boost. My plan for > now is to look carefully at what each of these programs were actually > doing while supposedly idle in the hope of developing a methodology from > that data. I don't have a preferred kernel scheduler; I brought it up because I think you're looking for more granularity of what's going on than you can get from looking outside of the kernel. Maybe you can get what you want from using something like oprofile and looking at sysfs kernel variables. -- Chris -- Chris Knadle [EMAIL PROTECTED] _______________________________________________ Mid-Hudson Valley Linux Users Group http://mhvlug.org http://mhvlug.org/cgi-bin/mailman/listinfo/mhvlug Upcoming Meetings (6pm - 8pm) MHVLS Auditorium Dec 5 - Open Source Show and Tell Jan 2 - TBD Feb 6 - DBUS Mar 5 - Setting up a platform-independent home/small office network using Linux
