[Red5devs] 0.5 Shaping up :)
steven.gong at gmail.com
Thu Jun 29 06:33:08 EDT 2006
On 6/29/06, Luke Hubbard <luke at codegent.com> wrote:
> Hi Guys,
> I thought I would drop a mail to the list to thank Steven for his work
> on the streaming code (which is looking ace, need a bit more time to
> take it all in), and Joachim for leading the charge with fixing bugs,
> adding features, and getting 0.5 ready. We spent a bit of time last
> night tracking down a rather odd intel mac bug, and generally watching
> buffer levels. For most things we seem to be down to zero buffer loss,
> which is excellent. Its now time to start testing the new streaming /
> bandwidth code, and fixing any bugs before 0.5 release.
> I have run some tests today on the Mac PPC (dual g5 2.5, 2gb).
> Here are my results.
> 1 Broadcaster, 1 Subscriber
> To spot lag I had to point it at a clock watch the second hand from a
> distance, and watch for a while. It was hard :) Which is great!
> 1 Broadcaster, 55 Subscribers ( On one pc )
> It worked ! If you hit f10 on a mac all the windows arrange in expose
> style.. And the videos are still running :). There was quite a bit of
> lag, but I attribute this more to the fp than the server. Server
> stayed around the 15-23% range cpu. Remember its running in eclipse
> ,without a load of ram, so this is good. Also we still have room for
> performance improvements. Profilers at the ready!
> In both tests the subscribers never disconnected. Well in the 55 one a
> couple of subscribers did drop off after 10 mins. Not sure why.
> However the broadcaster did seem to cut out after a while Normally 3-4
> mins. Im not sure if this has something to do with bandwidth control,
> or perhaps some unexpected packet. No errors seen in logs. Time seems
> to be variable but feels like it gets shorter after trying a few
> times. If the broadcaster is disconnected then reconnected, the
> streams start flowing again to subscribers.
Actually, the broadcaster doesn't use anything of the flow control now.
Anyway we should look into this issue.
Can we all run tests using trunk on our systems and report findings to
> list. Include pc type, speed, memory, number of subcribers, and java
> process cpu. Also if you get the bug with stopping broadcasting how
> - Luke
> Red5devs mailing list
> Red5devs at osflash.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Red5devs