There's a long and glorious history of debate, it seems, of the scalability of Asterisk. I've read of the recent steak-prize claim, but I was wondering: what sort of scale people are running out "in the wild" (as opposed to a benchmark lab setup) with Asterisk and Thirdlane?
I like the Thirdlane interface, but have concerns about it holding up under large-scale use (I'm looking to build a system that can support 15,000 users).
Any experiences folks are willing to share is appreciated.
Thanks!
Is there a reason I need
Is there a reason I need MTE? I'm not looking at doing multiple tenants. (I do understand spreading the load across multiple boxes.)
That's the $64,000 question
That's the $64,000 question for this project, but as a wild guess based on the trunk usage on our existing PBX, I'm thinking no more than 400. I guess I'm just wondering what'll happen if/when we become victims of our own success.
what kind of client is it
what kind of client is it that has 15k handsets but only 1/12 of a busy hour per user?
I never said 15K handsets -
I never said 15K handsets - just 15K users. We are a university, with approximately 15K faculty and staff. The plan is to give everyone an extension on the system, and then let them use the user portal to forward that extension to their cell phone, let it roll to voicemail, forward it off-campus, or whatever. If they want to buy a SIP set or download a softphone and register it, they can do that.
So I'll have 15K users defined, but I won't know ahead of time how many of them will be active, or what the usage pattern will be for those who are active. The 400 is a wild guess based on trunk activity for our existing PBX, but even that is pretty wild because I really don't know how relevant a comparison it is.
Hence why I was asking about existing large-scale installs so I could get a sense of where things would hit the wall.
there's other universities
there's other universities using asterisk
sam houston university
university of penn
Copiah- Lincoln Community College
some of them might know how many users per PBX they have running and what limits they ran across. Its mostly going to be an issue of hardware, concurrent calls, and the things you do (value adds) that stress the box. conference calling, recording calls, lots of voicemail; these are the things that will require more 'ass' in terms of hardware and also impose lower limits on concurrency. Since no one will need to know presence or place extension to extension calls, you could distribute users among more than one server and route the TN to the appropriate pbx while keeping the staff on their own pbx since they do do extension to extension calling.
I would not hesitate to go all out and get the HP DL380 G6 that have the quad-core + hyperthread + QuickPath processors in them (xeon 5530 or higher), 8gig of ram, and with that many users I would go straight for RAID-10 on SAS drives for the faster disk i/o. The other issue to contend with will be the user portal.
Thirdlane is a webmin portal and webmin was never designed to scale to a massive amount of concurrent access. But its a web portal for phone stuff. Chances exist that you may never have more than 3 people ever in there at one time. But if only 1% are in the portal; at 15k extensions, thats 150 concurrent web browsers. That's almost a bigger unknown than concurrent calls except that slow web access only makes for an annoyed user, has no impact on call quality.
you will need multiple MTE servers for that many users. However, with the right hardware, they can all connect up to a single asterisk gateway server so that your sip provider does not have to manage sending to/from several machines.