Gustav how many ASs do you see on your ntop instance? ASs were designed to be a few so I used a list. IMHO you have many ASs hence I need to change the datastructure.
Please let me know Thanks Luca On Mar 16, 2010, at 12:49 PM, Gustav Koller wrote: > On 3/15/10, Gary Gatten <[email protected]> wrote: >> Run in gdb to see what thread is doing what. I usually use top on just the >> ntop process and enable threads to see which thread(s) take the most time. >> Then with the thread I'd I can use gdb to see what those threads are doing. >> Or, *maybe* if -K and -t 5 is set ntop will write the thread name and I'd to >> the log on startup, but I don't recall for sure... > > I tracked this down to updateASTraffic() (pbuf.c), 80% of the CPU time > is spend in this function. I if I prevent updatePacketCount() from > calling updateASTraffic() everything works fine. Only a few packets > are dropped (0.1%). > > It looks like updateASTraffic() is called for every packet?!? > So for every packet ntop loops over the > myGlobals.device[actualDeviceId].asStats list. > After a few seconds running my asStats list had ~2000 entries. > With 248320 calles to updateASTraffic() the while(stats) loop in > updateASTraffic() already looped 10970927 times. > > Where are these asStats acutally used? IP/Summary/ASs works fine > without that updateASTraffic() call. I only could find a reference in > the rrd plugin. > > I guess we should move this to ntop-dev. > _______________________________________________ > Ntop mailing list > [email protected] > http://listgateway.unipi.it/mailman/listinfo/ntop --- If you can not measure it, you can not improve it - Lord Kelvin _______________________________________________ Ntop mailing list [email protected] http://listgateway.unipi.it/mailman/listinfo/ntop
