network

confessions of a pixel pusher technology

Just finished one phase of a bigger network job. The client probably did spend more money on engineering time than on what those ten network switches would cost. Yet, it was worth it. The guy I was working for is probably one of the most gifted engineers I have ever met. That of course was 90% of the reason of the success of the project. In the end we were able to get a gigE backbone in place, where there was none before. Despite the fact that the hardware existed, connections simply were done in such ways, that there were a couple of gigE islands (many as small as one switch) inside some very convoluted hundred base T soup. Valueable lessons I have learned:

– working with good people makes everything fun. Crawling up on racks @ 3am in the morning is perfectly OK under those circumstances, since the job will be finished

– the mess that accumulates over time in network topologies is huge. And damaging. Network switches are actually very brave in dealing with circular routes and all kind of crazy and unintended ways to connect things.

– snmp can be a great tool. It’s worth getting managed switches just for that. Apple’s Airport Base Stations require -of course- some extra care and work. All other network gear I used behaves very much alike. No matter which vendor. Of course Apple thinks they need to ‘think different’ even when they shouldn’t.

– Dell uses cross over serial, while Linksys uses straight ones. Xon/Xoff in Zterm on Mac does not work as advertised. Linksys latest srw2024 firmware will break the switch. It simple will no longer boot. (see comments) Linksys support is trying to help but is in general a clueless waste of time. Managed Linksys switches like the srw2048 or srw2024 I can not recommend.

– Hardware cost is in no relationship to the cost of configuring things. There are ample parameters to configure and tune your network. I doubt that the knowledge how to do this widely exists. I understand maybe 5% of what a 600 US$ managed switch can do and how it really works. Applying what happens with network to cars would mean that people drive around without tires. On their rims. It’s loud. Traction is lousy. So they buy a Ferrari (without tires) since the Porsche did not do so well (since it missed tires) and the rims were grinded down. And, yes, with a Toyota and some rubber around your RIMs you can be the fastest kid in town.

– crawling around racks I realized how much wire their used to be for Video and Audio. And just a couple of Cat5e cables can replace so much of it. What a concept that you need a different piece of copper depending on what kind of content would flow over it! Sure, realtime and all. But at what cost? Converter boxes, routers, distribution amplifier, patch panels, spliters. And then you are locked into, let’s say NTSC. It’s ridicolous, and much of it will go away. Writing a telegram was a great way to communicate. Once upon a time.

With managed switches, snmp and the dns database we had before we can now find the port / switch for every machine in the building. There are ample uses and applications for this. Looking at network traffic. Finding uptime. Hell, one could even track peoples work hours by that. Or at least give them access to it, so they can use their MAC address’ existence in the network as evidence for them being at work.