[Serusers] Scaling SER

Vitaly Nikolaev vnikolaev at intermedia.net
Wed Sep 20 18:03:45 CEST 2006


There are stateless or statefull Sip proxy operational modes, second is
what I use and it require same server that send request received
response.


Why do you need all this mess btw?

If you just have simple SER server it can handle up to A LOT of
customers (100k?), for redundancy you can use heartbeat with t_replicate
in ser.

If you use rtpproxy, then just use distributed rtpproxy, get few linux
boxes install rtpproxy on it and let ser/nathelper do
redundancy/failover between them for RTP traffic, all this feature
(except heartbeat) are exist in ser

But if you want to have different SERs in different locations, and have
customers distributed between them, you have to have "smart" something
that will route call between them (b2bua of course)





-----Original Message-----
From: Natambu Obleton [mailto:nobleton at fasttrackcomm.net] 
Sent: Wednesday, September 20, 2006 9:47 AM
To: 'Greger V. Teigre'
Cc: Vitaly Nikolaev ; serusers at iptel.org
Subject: RE: [Serusers] Scaling SER

I bet an Anycast setup with multiple servers having the same ip address
pointing to the same Mysql database would work. I have done it with mail
servers and dns servers.

http://en.wikipedia.org/wiki/Anycast#Reliability_of_anycast


Natambu Obleton
Network Engineer
FastTrack Communications
nobleton at fasttrackcomm.net
(970) 247-3366 office
(970) 247-2426 fax
 
-----Original Message-----
From: Greger V. Teigre [mailto:greger at teigre.com] 
Sent: Tuesday, September 19, 2006 11:55 PM
To: Natambu Obleton
Cc: 'Vitaly Nikolaev '; serusers at iptel.org
Subject: Re: [Serusers] Scaling SER

Actually it depends on the NAT implementation. There are four generally 
known/defined types (in STUN), but a lot of variants.  When we talk 
about the NAT problem, we always talk about the worst-case, which is a 
symmetric NAT, i.e. only packets from src ipA:portB will be allowed 
through and only if a previous packet has been going out to dst 
ipA:portB not longer ago than a defined time-out. Some NATs have timeout

of 60 minutes, others (could be) as low as 30 sec.
g-)

Natambu Obleton wrote:
> Ohh.. I thought the keepalives would allow other ip address, than the
one
> that originally opened the connections, to send in. 
>
> I guess that wouldn't be very secure would it.... :)
>
>
> Natambu Obleton
> Network Engineer
> FastTrack Communications
> nobleton at fasttrackcomm.net
> (970) 247-3366 office
> (970) 247-2426 fax
>  
>
> -----Original Message-----
> From: serusers-bounces at lists.iptel.org
> [mailto:serusers-bounces at lists.iptel.org] On Behalf Of Vitaly Nikolaev

> Sent: Tuesday, September 19, 2006 8:53 AM
> To: serusers at iptel.org
> Subject: RE: [Serusers] Scaling SER
>
> To extend Juha's info:
>
> Only SER server that just received SIP packet from SIP device (either
> register or keepalive) can send something  back in case of NAT
>
> For example you have SER1 and SER2, client A registered on SER1 and
> client B on SER2, when client B sends call to client A it will first
hit
> SER2 then SER2 to according to replicated location table will try to
> send it to client A straight and fail in case client A behind the nat
> (which is quite often in real world)
>
> Trick is to teach SER2 to forward call to SER1 when needed, it can be
> done when you use b2bua that looks to same location table for example
> (ser need to be taught to place its IP also in same table).
>
> Can not give you exact howto but can give (out of list) some
references
> to  commercial software if interested.
>
>
> -----Original Message-----
> From: serusers-bounces at lists.iptel.org
> [mailto:serusers-bounces at lists.iptel.org] On Behalf Of Juha Heinanen
> Sent: Tuesday, September 19, 2006 8:52 AM
> To: Andrey Kuprianov
> Cc: serusers at iptel.org
> Subject: Re: [Serusers] Scaling SER
>
> Andrey Kuprianov writes:
>
>  > mediapoxy'll handle it.
>
> mediaproxy does not forward sip requests.
>
> -- juha
> _______________________________________________
> Serusers mailing list
> Serusers at lists.iptel.org
> http://lists.iptel.org/mailman/listinfo/serusers
> _______________________________________________
> Serusers mailing list
> Serusers at lists.iptel.org
> http://lists.iptel.org/mailman/listinfo/serusers
>
> _______________________________________________
> Serusers mailing list
> Serusers at lists.iptel.org
> http://lists.iptel.org/mailman/listinfo/serusers
>
>
>   




More information about the sr-users mailing list