Writing

Feed Software, technology, sysadmin war stories, and more.

Sunday, July 3, 2011

How random is random, anyway?

I got to witness some strange things done on school networks. One time, the "network engineers" replaced the usual Bay Networks routers on a fractional T1 link with something that claimed to do compression. They figured that it would make things faster somehow.

While this was going on, some people from a nearby engineering company that made Ethernet testing and metering devices wanted to visit. Apparently, they had some new device on the way and wanted to try it on a "real" network. We had that crazy link with the compression and wanted to see what it would report.

So we went out there: the "network engineer", two tech types from the meter company, some kind of pointy-haired boss from that same company, and me. They had already plugged in their remote/helper box into one router, so we met up at the other end: at the school. There, in that room behind the school office, they ran their test.

It started slow. Then it got fast. Really fast. Ridiculously fast. Wow, they said. This thing must be working really well! Someone said it must be the compression. They said it was random, so that's not possible.

I asked one simple question: how often does it randomize the payload? Just at the beginning of the test or for every packet? They said, oh, yeah, it's random when we start the test. (In other words, they were sending the same payload over and over and over...)

I shook my head. Haven't you heard of compression dictionaries?

The pointy-haired boss just asked: how hard would it be to re-randomize it for every packet? They gave a typical engineer answer: two weeks!