« Teaching Complexity | Main | Utah NANOG »
April 10, 2003
Working in the lab
Tuesday morning we spent a few hours in the lab, testing some video Quality of Service designs. Well, that's what we planned to do. We had allocated 3 hours, and had 4 engineers working on the lab.
We were testing how QoS impacts H.323 traffic over a T1 (1.5Mb/sec), with various levels of traffic and different kinds of QoS mechanisms. The test itself wasn't that complex, just a couple of H.323 Polycom units with a couple of routers and a T1 in between.
It took the four of us roughly 2 hours to get the test set up (not counting the several days Mike Downie had already spent scrounging up equipment, cobbling together enough pieces to do basic testing, and getting them installed in the rack). Between re-configuring the Polycoms to configuring routers to getting laptops connected and loaded with test software, we used up almost the entire lab time. As always happens with technology, things didn't work at first, and required a fair amount of debugging.
We tried using WANkiller, a piece of software that can (supposedly) generate gobs of traffic to saturate a circuit. For some reason, the version we had would only generate 40% of the capacity of the circuit. So we decided to get iperf set up. More time getting laptops back on the Internet and getting that software downloaded and installed.
Finally we could start testing, with less than an hour left. Our initial tests were pretty interesting. We could not break the video or audio, other than brief freezing. We were running the T1 at 100% in both directions, but H.323 was working just fine. We spent a few minutes trying different settings on QoS, the software and the Polycoms and were able to break video in some cases, and keep it clean in others. These tests generated as many questions as they answered.
Unfortunately, everyone had other responsibilities for the remainder of the day and week. We had to dismantle the test, so we could use our laptops for normal work, and let other people use the Polycoms for other lab work. The hour of actual testing we got to do gave us some direction, but it was by no means scientific, and there was no time for good documentation or in-depth analysis. Hopefully when we follow-up, we will remember the results of our first day.
I really enjoy working in the lab. I think that's where many new, good ideas can be generated, from hands-on tinkering. I'm also finding it's critical to be able to lab and test projects before they are deployed. But if each project takes hours and days to set up, just to squeeze out an hour or two of testing, I won't get to spend much time in the lab. Gotta figure out a better way.
Posted by pete at April 10, 2003 12:36 AM
Trackback Pings
TrackBack URL for this entry: