afc@shibaya.lonestar.org (Augustine Cano) (05/23/90)
In article <1990May22.190038.5217@bagend.uucp> jan@bilbo.uucp (Jan Isley) writes: >There was some discussion many moons ago about what the numbers mean >for load average with sysinfo. Perhaps like everyone else, I thought >that a load average of 1.0 meant the system was maxed out. When I >noticed numbers like 1.05 I did not think too much about it. However, >as I type this, I am running the byte benchmarks posted to comp.sources >a few weeks ago. I will post the results if it ever finishes. It has it will finish in about 6.5 hours. I also ran them (at night.) One of the tests (repeated three times) exhausted the system resources. The last two times, there were failed execs galore. A couple of other things failed due to minor modifications that had to be made. One of these days, I'll re-run the benchmarks. >been running for many hours now with the usual background stuff going >on, reading news, a few megabytes of news coming in and going out, etc. > >The load average just read: 5.36 4.95 3.90. Seems a bit high? > >Someone want to tell us again what these numbers really mean? They indicate (if I recall correctly) the load average in the last minute, in the last 5 minutes and in the last 15 minutes. Why they are so high, I don't know (would someone care to enlighten me?) I have also noticed that the load average numbers generated by sysinfo seem to get stuck periodically at different values, even when there is nothing going on. Does this happen to other people out there? I thought this was related to the previous buggy kernel, but I'm now running 3.51m and it still happens. The most common values to get stuck at are 0.04, 0.1[34], 0.3[0-4] and sometimes 0.4?. Is this a bug in loadavg? is there a fixed version out there? > >jan > -- >jan@bagend | {..gatech..}!bagend!jan | (404)434-1335 = voice@home > >Signature, we don't need no stinkin signature. Augustine Cano afc@shibaya.lonestar.org
wayne@csri.toronto.edu (Wayne Hayes) (05/26/90)
In article <1990May23.160031.15723@shibaya.lonestar.org> afc@shibaya.lonestar.org (Augustine Cano) writes: >In article <1990May22.190038.5217@bagend.uucp> jan@bilbo.uucp (Jan Isley) writes: >>been running for many hours now with the usual background stuff going >>on, reading news, a few megabytes of news coming in and going out, etc. >> >>The load average just read: 5.36 4.95 3.90. Seems a bit high? >> >>Someone want to tell us again what these numbers really mean? > >They indicate (if I recall correctly) the load average in the last minute, in >the last 5 minutes and in the last 15 minutes. Why they are so high, I don't >know (would someone care to enlighten me?) Well, I don't have UNIX on my PC (just reading this group for info), but on the Suns at school, the load average means "how many programs are actively competing for CPU time?" So, one CPU hog and nothing else gives a load of exactly 1.00; if you run 2 CPU hogs (and nothing else) for 15 minutes and then sneak a load average check (`uptime` on our Suns), it should read something like "2.10 2.02 2.00" (the 1 minute and 5 minute loads are affected by the load-checking command itself). If all your processes are sleeping, it's 0.00. On a day that a big ass't is due at school here, I've seen the load average upwards of 20, and even (rarely) near 40. And when someone in the Unix and C class doing the "write your own shell" ass't accidentally starts a fork() inside an infinite loop, I've seen the average skyrocket to over 200. Of course the uptime command itself takes about 20 minutes in this case, and the system is completely unusable. Hope this helps ... -- Mathematics: That branch of Human Thought which takes a finite set of trivial axioms and maps them to a countably infinite set of unintuitive theorems. Wayne Hayes INTERNET: wayne@csri.utoronto.ca CompuServe: 72401,3525