Hello, On Fri, Nov 19, 2004 at 03:04:52PM -0500, Joey Hess wrote: > Attached is the raw data. I haven't processed it since I don't have > matplotlib installed. > > The system is a completly stock sarge desktop autoinstalled this > morning, running on a fairly beefy proliant server. To simplify things, > I booted it with a uniprocessor kernel. Given that total boot time is > only 24 seconds even with top spinning, the 0.2 second sample interval > may be too high to catch all processes. Ok, with your new data I got it to work. The result (0.1 second sample interval) is at http://seehuhn.de/comp/bootlog3.png Boot time on this machine after init is started is approx 23 seconds. Quite impressive. And 7 of these 23 seconds seem to be spent waiting for the DHCP-server. The most plausible candidates during seconds 3-5 for further optimisation might be the depmod call and the 3 seconds sleep caused by the /etc/rcS.d/S4... (name truncated) script from seconds 8 to 10. Minor issues: 1) Either the top output format changed between sarge and sid, or it is other influences. Your data did not include the I/O wait times, so they are omitted in the diagram. My top headers look like this: top - 12:55:35 up 0 min, 0 users, load average: 0.15, 0.03, 0.01 Tasks: 19 total, 2 running, 17 sleeping, 0 stopped, 0 zombie Cpu(s): 2.0% us, 8.6% sy, 0.0% ni, 65.6% id, 23.8% wa, 0.0% hi, 0.0% si while yours look are top - 20:09:54 up 0 min, 0 users, load average: 0.00, 0.00, 0.00 Tasks: 12 total, 1 running, 11 sleeping, 0 stopped, 0 zombie Cpu(s): 6.9% user, 10.9% system, 0.0% nice, 82.3% idle 2) Top has this annoying habit of truncating the output according to the screen width. Yours seems to be 80 columns which causes some of the process names to be truncated. If somebody can coerce top into providing the missing information I am very willing to have another go at the data. All the best, Jochen -- http://seehuhn.de/
Attachment:
signature.asc
Description: Digital signature