Some other live streaming tests

[Gavriloaie Eugen-Andrei]

Environment:
- FMLE running on windows pumping data at ~800Kbits/s (high vp6)
- rtmpd running on Mac OS X (2.66 GHz/4GB ram)
- flazr (more details on this great tool on later posts) stressing the server with 200 concurrent playback connections running on the same Mac OS X laptop

Numbers:
- rtmpd ~5% CPU
- rtmpd 36.4MB ram
- total throughput at ~20MB/s
- flazr instances ~7% CPU for each instance (100 connections per instance, 2 instances)
- flazr ~310MB ram per instance

Also connected a real flash player to rtmpd and see how is the playback. Crisp, clear and cursive.

Another facts on rtmpd:
- compiled in debug mode
- IOBuffer asserts where in place for each memory operation. This ads almost 20% overhead, so if I remove them the server should stay even lower than 5% CPU
- other "critical places" asserts were in place.

Comments? Suggestions?

Best regards,
Andrei

P.S.
Also attached a screenshoot

[Ondrej Mikle]

On Nov 27, 4:56 pm, Gavriloaie Eugen-Andrei <crtmpser...@gmail.com>
wrote:

>  Screen shot 2009-11-27 at 5.45.48 PM.png
> 310KViewDownload

Excellent! I have to say rtmpd is very good when it comes to
performance/resource consumption.

I've been testing mostly recording, here are some results:

Since rev 851 recording is very stable, haven't had a single crash.
Repeatedly I get the following:

Numbers:
- 30-40 flash upload clients (a bit tedious, but I haven't found an
usable command-line rtmp publish client)
- 0.5 - 4 MB/s (depending on encoding parameters and network)
- around 0.2%-1% CPU (with occasional 2-5% peaks) and 7-8 MB RAM on
Gentoo Linux amd64 (2 GHz/2 GB RAM)
- also compiled in debug mode

Suggestions: if possible, try also testing with remote server (to
account for network latencies)

Regards,
Ondrej

[crtmpserver]

I will definitely try to stress the server on the wire. I'll come back
with some results.

Cheers,
Andrei