> [EMAIL PROTECTED] said: > > I am having real problems with what appears to be scheduling issues, > > and I am running out of ideas... > > Why do you think they are scheduling issues?
Thats a good question! More or less just speculation. If someone does something CPU heavy, and I nice them to 20, it doesnt seem to have any impact on their scheduling (gut measurement not benchmarked). Everything seems as interactive as before. This is a shame as I really want to slow heavy users down. Course it is hard to tell whats really happening, as I know that if there is spare cycles available then even nice 20 should still be scheduled. > Describe "die for a while". If it is a performance drop which goes away, I > would have a vmstat 3 (or 5 or 10, whatever is the relevant time scale) > running > and look for differences between the good performance periods and the bad > ones. loadavg looks not too bad in general. When I forgot to disable anacron in the UMLs it reached 4.0, but in general it is about 0.6. Its when it is above 1.0 I was looking for a way of crippling heavy CPU users... Students are mostly CPU-friendly, but the crazy ones who want to do "find / > /dev/null" to see what my system will do is the ones I need to watch. Wierdly my main server (the gateway machine plus other things) has a loadavg of 0.8-1.4 (its dual processor) but has an idle time of 95%. This is a production system, and my peak day is Wednesday. Any suggestions welcome, and I will implement them and get the students to test it next week! I really want to make this system as safe as possible, as we have plenty of periods where the system is idle and I want to open up the site for public free usage when the system is idle-ish. We have integrated with the machines interactive tutorials which log in and test to see if task objectives have been completed, and use this as a way to teach unix skills and system administration. I just dont want the internet clown community to use the machines for parallel processing or recompiling their kernels, and making other UMLs on that machine slow to a crawl. My only other approach is a CPU quota approach (100,000 jiffies per week?) but it seems rather arbitrary... Thanks for replying. G. ------------------------------------------------------- SF email is sponsored by - The IT Product Guide Read honest & candid reviews on hundreds of IT Products from real users. Discover which products truly live up to the hype. Start reading now. http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click _______________________________________________ User-mode-linux-user mailing list User-mode-linux-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/user-mode-linux-user
