Defacto QoS Revisited

Last week I made an off-the-cuff observation about QoS that turns out to be only partly true.  In some private correspondence, David Reed pointed out that, yes, on a congested link UDP traffic will take precedence over TCP traffic for the available bandwidth but this doesn't help VoIP traffic.  When congestion occurs, the router queue in front of the congested link fills up and introduces significant delays.  Upon reflection, this is obvious.  I'll have to be a little more careful in future posts...  Meanwhile, some quick experiments over the weekend quantified exactly what happens with the specific setup at my house.

I have Comcast Internet service in Massachusetts with a Motorola SB5100 cable modem that's connected to my home LAN via a Netgear Cable/DSL Wireless Router MR814 v2.  The best results I've obtained with a bandwidth speed test (on Sunday morning at 6:30am) are 3850 Kbps down and 357 Kbps up.  However, when I ran the tests below it was Sunday afternoon and a typical bandwidth test yielded 2847/345 Kbps.

To test my upstream performance under load, I wanted a nearby site with no other bottlenecks besides my upstream link.  MIT is such a nearby site for me (both physically and in a Internet sense), as MIT peers with Comcast in the Boston area.  Traceroute shows one router, 24.218.0.194 in Cambridge Massachusetts, between Comcast in Woburn Massachusetts and the MIT backbone.

Sure enough, when I Ping MIT repeatedly, the round trip times average 15 ms (minimum 11ms max 33 ms):

Repeating the Ping sequence while simultaneously sending an email with 10 MB of attached photos and I get a wide variation from 13 ms to 349 ms with an average of 171 ms round trip delay plus 1% packet loss:

This is not completely fatal for VoIP but, after de-jittering, that's 360 ms to 400 ms of added delay -- something that's very noticeable on a voice call.

So my apologies for the earlier post.  Meanwhile, I've done a little research on the web, especially on gaming forums, as gamers are extremely interested in low latency.  As a result, I've ordered a Linksys WRT54G router from Amazon which I plan to try with QoS firmware from Sveasoft.  I'll report the results when I get them.

 

Comments

[Tim Garthwaite]

Hi, Brough!

I use the WRT54G at home. You don't have to pay Sveasoft for their firmware, as it was once released under GPL and they simply stopped distribution of it. Here's a good site to grab that free version, as well as a few community-enhanced versions.

http://wrt54g.thermoman.de/

Tim