• […] bookmarks tagged concurrent Read my words – 3500 concurrent channels with Aste… saved by 3 others     Phillychesse bookmarked on 02/18/09 | […]

  • Hi! I am trying to setup a high availability asterisk cluster on ec2 to handle 150 concurrent calls and I’m wondering the best way to go about it. Through our earlier conversation and research I recognize that the Asterisk Meetme functionality does not work on ec2 but this is not an important feature to me. The cluster needs to be scalable because although right now I am only planning to accommodate 150 concurrent calls during peak time, I desire to be able to handle at least double that with the addition of a couple of more asterisk machines. It seems like some people have done it using heartbeat but from what I’ve read it seems like that is more for redundancy than high availability. Also I’ve read some articles that seemed to use database replication in which one machine was responsible for keeping the master database and each asterisk node kept a local copy of the database that it read from but did writes to the master (which trickled down to the local dbs) and DUNDI was used in this installation.From a load balancing standpoint is openSer the best option with the dispatcher module?

    Here is the ultimate plan: To have an ATA device encode and decode calls using g729. Use openSer to direct traffic and handle registrations. So for instance if someone wanted to make a call from the ATA to the PSTN, OpenSer would help establish the invitation process between the ATA and pstn gateway and would not be included after that. Asterisk would only be used for voicemail and other features for the person on the ATA. I’m not sure if this is even possible put I would like for normal calls (from ATA to ..say someone on the PSTN) to not route their RTP traffic through Asterisk.

    I’m sorry if this confusing. I’ve been furiously reading to try to understand but I’m sure I’m not close to being there. I would appreciate any help at all.
    Blake McKeeby

    February 21, 2009
    • Hi Blake,

      Sounds like your setup is fairly common from what I can tell. It is true that Heartbeat is mainly used to High-Availability. Load balancing requires the use of either OpenSER (or a similar software) or a vendor Session Border Controller. Bear in mind that the NAT based network environment of Amazon EC2 may pose some issues.
      In terms of configurations, there is no single answer here, as it requires a proper analysis of your provisional business model. Feel free to contact me by email and I’ll see how I can assist you on your quest.

      February 21, 2009
  • If H.323 would suffice you may be interested to know that a single instance of GNUGK easily handles 60,000 call setups per hour with a load average of 0.1, and that’s with the Postgres server on the same machine. That hardware is getting old now: 2GHz AMD64 X2, 2GB DDR RAM and mdraid mirroring across 2 SATA drives.

    February 21, 2009
    • Hi Stavros,

      I’m not sure if you recall, but I’m well versed in GnuGK. Back in 2003 I used GnuGK+FreeRadius+MySQL to traverse over 25 Million minutes a month. I have to admit that the idea of running a cluster of GnuGK systems using Amazon EC2 is interesting, however, the nature of H323 and its incompatibility with NAT and the fact that Amazon EC2 is based on NAT networks, I would say that it poses an interesting issue.
      In any case, GnuGK was optimally built to handle call setup requests, as it’s a GateKeeper, very much like a SIP proxy or SIP Session Border Controller, while Asterisk isn’t really built for that kind of function.

      February 21, 2009

Leave a comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.