Earlier, I used webbench from my laptop (webbench -c 100 -t -1 (100 processes)) to stress-test the gateway box running Apache2.
Running webbench with these options, system load on the gateway box would rise to ~120 and the system felt a little sluggish when using the vi editor for example.
After having made the switch to SCHED_ULE + PREEMPTION the box feels snappier even at the same system load.
However, I use the munin and munin-node ports on my gateway box and I do admit that the munin-node process (perl) refused to update any of the statistix while the system was running under a >100 load.
Only after having killed the webbench stress-test the statistix were gradually updated (or at least, the generated png's were).
This is something I need to take a closer look at.
posted by .marco. @ 22:06
0 Comments:
Post a Comment
<< Home