I know, I was just saying for when
he starts doing some load testing (v 0.5 or 0.6) that I would be interested in
these numbers as im sure most of the load testing will be geared towards
streaming media. Just thought I would ask as Rob said “If there are any
particular scenarios you would like me to test”.
Jim
From:
[EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Luke Hubbard
([EMAIL PROTECTED])
Sent: 21 April 2006 11:48
To: [email protected]
Subject: Re: [Red5] load testing
Its been said before, but
I will say it again. The 0.4 release is a transitional release which
intergrates the (very new) api. Its meant to work for people to start
developing with. Not designed to go anywhere near production. I know some will,
but we have to give the warning anyway. The next release will focus on
refactoring streaming / messaging code. So there isnt much point profiling 0.4
or pushing to the limits as things are likely to change. We would however
welcome people load testing 0.5, esp before the release, so we can see where to
make improvements. Should we start a new thread on how best to put together a
load testing framework for red5?
-- Luke
On 4/21/06, Jim Tann
<[EMAIL PROTECTED]> wrote:
I am interested in how many users can be connected to a shared object,
and to a red5 server in total without streaming media (just shared
object connections & some AMF)
Jim
-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED]]
On
Behalf Of Rob Terrell
Sent: 19 April 2006 17:14
To: [email protected]
Subject: Re: [Red5] load testing
My company has heavy-duty FCS use, both on single servers and using
Flash CDNs like VitalStream and Akamai. We regularly QA our servers
(and our streaming partners) against 500, 1000, 5000 and more RMTP
streams -- we've provisioned events for as many as 40,000 streams.
Once .4 is out, I'm planning on using our testing center to do a
reality check on Red5. If there are any particular scenarios you
would like me to test, let me know.
And John is right -- the single point of failure is a big problem. I
have spent some time working on a poor man's clustering setup for
Red5, although with the API in flux it's not a serious effort right
now. Definitely it will be a major component for me once the API
settles down.
On Apr 19, 2006, at 10:46 AM, John Grden wrote:
> I've often heard numbers like 1000 concurrent connections on a
> single box being a common max point, but I'm sure it can change per
> situation based on what you're doing. The issue is quality of
> experience with that many users obviously. *Can* a server do
> 2500? I doubt it, but I could be totaly wrong. And
even if it
> could, it'd bprobably be a miserable experience.
>
> Plus, that's just a single point of failure waiting to happen -
> what about failover/redundancy etc? I've been down this road on
a
> major scale and you have to have something planned for load
> balancing the community (hardware AND software IE: loadbalancing
> via the switch that handles your traffic etc), redundancy etc.
_______________________________________________
Red5 mailing list
[email protected]
http://osflash.org/mailman/listinfo/red5_osflash.org
_______________________________________________
Red5 mailing list
[email protected]
http://osflash.org/mailman/listinfo/red5_osflash.org
|
_______________________________________________
Red5 mailing list
[email protected]
http://osflash.org/mailman/listinfo/red5_osflash.org
