In message <20030519224139.GF29561 at amphibian.dyndns.org>, Toad <toad at amphibian.dyndns.org> writes >On Sun, May 18, 2003 at 05:14:33PM +0100, Roger Hayter wrote: >> I have observed that with my particular setup the routine use of nice >> with Freenet in Linux is not helpful. It leads to load averages of 1.5 >> to 4 instead of 0.5 to 1.5 using standard priority. >> >> Linux 2.4.19. Duron 1.2GHz, 256MB RAM, two nodes, one standard using >> 596 and one with a low number of threads using 6034. Limited bandwidth >> from 2.5kB/s 10kB/s. Network load usually between 250 and 900. Waiting >> to send 1MB or so only about half the time. Is there a diagnostic metric >> for time spent waiting for bandwidth? Sun 1.4.1 or 1.4.2 beta. >> >> With nice -10 there is much more context switching and paging, > >Be very careful here. What priority are you setting it to, 10 or -10?
Point taken, typo, "lower priority by 10 units as in the default start-freenet.sh" is what I meant. > >> presumably because threads are not being allowed to finish what they are >> doing. In both cases, CPU is "idle" for 40-80% of time. Is there a way >> of measuring "voluntary" and "involuntary" context switching in Linux? I >> presume a thread waiting for bandwidth in some way will voluntarily >> relinquish control of the CPU. >> -- >> Roger Hayter -- Roger Hayter _______________________________________________ devl mailing list devl at freenetproject.org http://hawk.freenetproject.org:8080/cgi-bin/mailman/listinfo/devl
