Client connects via T1 line to provider. VOIP server is on provider's local network.
Provider: Cisco router <- T1 -> Cisco router <- m0n0wall -> customer LAN
I set up four pipes (2 down, 2 up) on a client firewall. The idea was to segregate fully a portion of the bandwidth for VOIP.
For the VOIP traffic, I then set up rules which used the WAN interface and had the destination as the remote VOIP server (e.g. 20.21.22.23). The rule was for any protocol, any source, any port.
The other rules ran into normal queues, and those queues used the other two up/down pipes for the remaining portion of the bandwidth.
In practice, this didn't work. When looking at the traffic graph, it appeared that the max available bandwidth was what I set the generic traffic to use. I even got on a voice call, saw the in/out traffic graph stay steady around 100K and had the user do a speed test. I could see the peak go up to about what the max traffic was.
Either way, there were still some voice quality issues when there shouldn't have been.
What is the proper syntax to add a rule so that ANY traffic to the VOIP server 20.21.22.23 will use the VOIP pipe up and VOIP pipe down?
Thanks.
ps - Is it generally held to be true that using the generic shaping and sharing of all bandwidth will introduce latency due to the use of queues? Some of what I've read suggests that using queues is not preferred and that VOIP traffic should simply be put directly into its own pipe.
Real world feedback and configuration examples would be appreciated. Many of the archived items no longer have screenshots available, and the WMV screencast file won't play on my Mac under Windows Media Player 9, WMV, Flip4Mac/QuickTime.
|