[Red5devs] Question Regarding Load Testing RTMP

[mabrek]

On Mon, Sep 1, 2008 at 1:48 PM, Scotty G. <scottyg at gmail.com> wrote:

> Thanks to everyone for this information. When 3k users is mentioned, what
> does that scenario look like? In other words, is that 3k users sending
> RPC-like message requests to Red5, which then instantly responds (i.e. an
> echo service)? Or is every message sent from the RTMP client(s) to Red5
> pushed to all other connected clients?
>

My scenario was live streaming. One stream from client camera up, 3k users
watching it.


> To give more context, for my AIR client, I'd like to make RPC-style
> requests over RTMP for the basics. These can be message-based and
> asynchronous, but there's still an initial request with an expected
> response. While this is happening, other events in the system can occur,
> which will then result in notifications that need to pushed down to specific
> clients (i.e. 20 users have been waiting for a server resource to become
> available; the resources just freed-up and it's time to notify everyone in
> the queue). To properly load test my system, I'll need to generate X
> RPC-style load from the AIR/RTMPClient clients, while pushing Y event load
> from the back-end. How would you set-up a scenario like this?
>

If I were to perform such test I would make Jmeter probe based on RTMPClient
that sends one message (make "invoke") and waits for answer. Than I would
play with Jmeter settings and test plans. Beware, single Jmeter instance has
GC troubles running more than 500 RTMPClients
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://osflash.org/pipermail/red5devs_osflash.org/attachments/20080901/b0b92bde/attachment.html