Well, in our company we just lease all equipment. Yes granted we try hard to keep the cost down, we figure the equipment cost is the least of our worries. Now that doesn't mean we'd spend millions on a softswitch, we're still a small company. :-) A layer7 switch, if you compare the price to a softswitch, isn't that much money. The only reason we would consider a hardware solution though at this point is because there is literally nothing (free) software wise out there.
What we're really wanting it for is Asterisk, since asterisk doesn't have the ability to do SRV lookups the only other alternative is butchering the dialplan/making a script to perform redundancy to peer with SER. With SER I was assuming use of SRV records otherwise, and for our registered users we would again use SRV records (UA compatibility pending :D) and hope t_replicate can perform adequately enough.
We have an SRV bounty for Asterisk, it isn't publicly posted, if anyone is interested ;-) No one stepped up to the plate yet in the ast lists.
Matt
-----Original Message----- From: Greger V. Teigre [mailto:greger@teigre.com] Sent: Tuesday, April 12, 2005 11:55 PM To: Matt Schulte; serusers@lists.iptel.org Subject: Re: [Serusers] Load Balancing via UltraMonkey/ldirectord
So, you found the budget? Or was the pain just big enough... Let's see what we can do. It will take some time, I assume, so meanwhile... g-) Matt Schulte wrote:
Heheh, we may actually test one of those F5network switches...i f we don't come up with a fairly painless, bugfree, and most importantly supportable solution. :-) I cannot code therefore I would be useless, but I can safely say learning C is on my to-do list. We may however be
willing to contribute to a bounty, at the very least I'd be more than happy to test :D
Matt
-----Original Message----- From: Greger V. Teigre [mailto:greger@teigre.com] Sent: Tuesday, April 12, 2005 5:29 AM To: Matt Schulte; serusers@lists.iptel.org Subject: Re: [Serusers] Load Balancing via UltraMonkey/ldirectord
:-) Yes, layer 7 switch is of course nice. But then again, you need to
make sure that you can price your services where people will buy them...
I wonder whether it is possible to gather some people interested in this and get something started on the development side. AFAIK, LVS struggles with
other UDP services too, so a ipvs UDP content analyzer would probably be of interest. I looked at the source code and I think the most difficult thing would be to extend the ipvs framework to allow a module to peak into the
packet (and not only the header). I don't know what kind of performance penalties you get either.
I have seen several people being willing to sponsor development. We could hire somebody at http://www.rentacoder.com/ ;-) g-)
Matt Schulte wrote:
Yah I noticed the other post after I posted mine, I don't see how it would easily be possible to address the sticky issue. It would require
making a SIP aware proxy of sorts, which is a bit out of my abilities.
Has anyone been able to address this issue? Of course a layer7 switch
would do wonders and eliminate the need for all this, but who has that
money laying around :D
I've done a little research (google) and noticed people mentioning it
when talking about LVS, one guy said he was going to write a module but posted nothing more. That would be pretty slick.
Matt
-----Original Message----- From: Greger V. Teigre [mailto:greger@teigre.com] Sent: Monday, April 11, 2005 3:04 AM To: Matt Schulte; serusers@lists.iptel.org Subject: Re: [Serusers] Load Balancing via UltraMonkey/ldirectord
If you see another thread (using the rather intuitive subject: Re: [Serusers] more usrloc synchronization), you will see discussions on using LVS in general. AFAIK, which high availability solution to use for LVS, is more based on your personal preferences, UltraMonkey is probably a safe choice. Anyway, you will need to address the "stickiness" issue. g-)
Matt Schulte wrote:
Has anyone attempted to load balance SER using Ultramonkey/ldirectord?
I've noticed all it does is pretty much NAT and send requests accordingly, the trick I guess would be the NAT part. If the SIP headers = myself, would there really be any issues? One problem I can
foresee is the possibility that loose routing would hit the wrong server. Just wanted to ask around before I wasted time trying it out
for myself :-) Thanks
Matt
Serusers mailing list serusers@lists.iptel.org http://lists.iptel.org/mailman/listinfo/serusers