[Asterisk-Users] Scalability of chan_oh323

[Michael Manousos]

Alistair Cunningham wrote:
> Michael Manousos wrote:
> 
>>
>> Alistair Cunningham wrote:
>>
>>> I have a customer who wants to do large volumes of H.323 to H.323 
>>> hairpinning. We haven't tested this scenario for large volumes 
>>> before; maybe someone on asterisk-users has.
>>>
>>> If they buy a top of the line PC, how many concurrent calls are we 
>>> likely to get? Routing logic will be simple, the machine won't be 
>>> doing anything else, and let's assume no transcoding for now.
>>>
>>> We're not looking for an exact figure at this point, just a rough 
>>> estimate for cost / benefit of Asterisk versus a proprietary system.
>>>
>>
>> Currently, without transcoding, you can get maximum 100 simultaneous
>> H.323 channels per box. With the next release of asterisk-oh323 this
>> number will be raised to ~180 channels. After that, major optimizations
>> at the OpenH323 RTP/jitter buffer code are required to push this number
>> up.
>>
>> Michael.
> 
> 
> That's a shame; my customer probably needs 400 to 500 channels (200 to 
> 250 calls).

I know but the performance bound is set by OpenH323. Also, an additional
note, these numbers are almost the same even when transcoding
(G.729 <-> G.711) is used.

> 
> Does anyone have experience of GNU Gatekeeper in proxy mode? Any idea of 
> what load it can handle?
> 

GNUGK in full proxy mode (all signaling and RTP through it) has a
similar upper limit.

Michael.