Tasks: 276 total, 1 running, 223 sleeping, 0 stopped, 0 zombie I haven't measured it but I can say, without taking the measurements, that v1.32 will not run my 8 cameras even if I do the video suggestion. A suggestion had been made that allowed_media_types=video should be added to rstp command in the video description. Maybe functionality was moved between the two processes?!Īnother assurtion was that zma was taking performance hits because sound data was considered in the new zma. But here we see just zma taking the performance hit and zmc is actually reduced. If, indeed, this was the case, wouldn't the expectation be that the performance hit occurs across both zmc and zma. Now lets consider the assertion that 18.04 security updates are accounting for the performance degradation. zmc, on the other hand, is improved somewhat. V1.32: zma utime 282 stime 0 <- 4 times moreĬonclusion: in the new release, zma is hogging the cpu by a factor of about 4 compared to the old release. I run the script twice just to see if the measurements are consistent when taken over different times.
# now (AGAIN) read proc data for zma and zmc so we can compute the cputime # now use top to report on user top -u www-data -b -n2 -d $waitTime WaitTime=10 #we will wait 10 seconds between measurements
#CPU UTIME STIME CODE#
It turns out system time is negligible - so this is all application code hogging the cpu.Ĭode: Select all #manually determine the process ids of zma and zmc Reading proc data is a the better way to go because we get to see if something funky is going on during system time of the process. It does appear that top's time+ is utime+stime (and perhaps dead children cpu times but that doesn't apply here). Now we compare to output of top command which reports time+. Read this file twice over a time period and you now have the amount of cpu (system and user) time consumed by the process. when a process runs it picks up clock ticks in utime versus stime depending whether the context of the process is user or system. Starting at field 14 of the file we have utime, stime, cutime, cstime measured in 100ths of seconds. Process performance data is available via /proc/$pid/stat where $pid is the process id. Okee doke: Nice architectural diagram over at /en/stable/userguide/components.html which shows each camera video feed goes to a zmc (camera) process that then feeds a zma (analyzer) process. zmaudit, reported in other threads as a performance hog, is turned off on both systems! I have the zoneminder camera set to record. I flip between hard drives to do experiments where one system is v1.30.4 /16.04.06 (zoneminder/ubuntu) and other is v1.32.3 / 18.04. My experimental setup: Hardware is i7-5600u with hard drive and 16gb ram and a single hikvision 2032 camera running 4 frames / second at 1920x1080. I'm finding zma (analyzer) consumes about 4 times the processor time in 1.32 compared to 1.30.
0 kommentar(er)
