Community
Participate
Working Groups
This ticket is to track work to ensure asterisk.eclipse.org receives sufficient bandwidth. Symptoms noticed: 1) ssh sessions are lost for no reason after mere seconds or at most a few minutes 2) Duplicate packets detected 3) Very poor sound quality and dropped connections, particularly the week of February 6th Denis and I talked this over in the hall. Our hypothesis for #3 is that high downloads are causing our bandwidth to be throttled. Throttled network and QoS is crowding out asterisk (which uses UDP). A previous discussion of ours, we hypothesized that #1 & #2 may believed to be related to a quirky routing issue between the .206 and .209 subnets. Once this issue is resolved, we can roll out Asterisk to replace our existing conference bridge system (and save lots of money)
I've completely refactored our QoS rules into four classes of traffic: 1. realtime 2. guaranteed 3. default 4. bulk Whereas before a bunch of non-downloads traffic would have to fight for the #1 spot, now only Asterisk and some network control traffic is in the realtime class. "guaranteed" contains other important traffic, including wiki, bugs, git/cvs/svn, and just about everything except downloads. We just held a test call, and there was hardly any noticeable jitter despite our pipe being heavily saturated. During the call I cut our bandwidth by about 30 Mbps and there was no noticeable jitter either. I won't close this as fixed just yet, but I'm willing to bet money that it is.
I think this is done.
