Hi,
On May 3, 2007, at 3:07 PM, Luke Hubbard wrote:
> Doh, it was right there..
>
> <snip>
>
> Ok, so the problem is with live not vod.
> If it doing any disk IO? recording?
Recording functionality is still included (as it was in oflaDemo),
but is not used currently. There's hardly any disk IO, looking at
iostat -k:
Device: tps kB_read/s kB_wrtn/s kB_read kB_wrtn
sda 0.81 0.18 7.51 107087 4535668
sdb 0.81 0.14 7.51 85520 4535668
md0 0.39 0.03 1.53 18453 922840
md1 0.06 0.00 0.22 1325 132040
md2 0.20 0.16 0.78 98121 468556
md3 0.64 0.12 2.55 70533 1538016
md4 0.01 0.00 0.06 1893 34184
I've also been doing some basic CPU usage profiling with Sun's hprof,
cpu=times gives this output:
rank self accum count trace method
1 34.99% 34.99% 7043 336190
java.util.concurrent.locks.LockSupport.park
2 30.27% 65.26% 6617 336361
java.util.concurrent.locks.LockSupport.park
3 6.43% 71.69% 2270 330107 sun.nio.ch.PollArrayWrapper.poll
4 3.26% 74.94% 80 303356 java.lang.Object.wait
As I'm not sure whether the profiling also counts time spent waiting
in a blocking state or how java's threading works internally (backed
by native threading libraries?) this output is either normal or
indicative of a threading problem. In that case I'll have a look at
trunk and play around a bit with the new options.
Maurits Dijkstra
_______________________________________________
Red5 mailing list
[email protected]
http://osflash.org/mailman/listinfo/red5_osflash.org
