sean's place

Bandwidth

My IRC server has become one of my more consuming interests for the last few weeks. From nearly the first day (the interim being to instrument what follows), I’ve been tracking statistics from the aggregate bandwidth usage for the machine (IRC server and all other services combined), the number of clients on the IRC server, and so forth.

The server itself has found its pattern of holding between 650 and 800 users at any given time. It’s also taken to hubbing most of the network in anticipation of the pending delink of our current hub.

Throughout all these changes, I expected to see bandwidth usage increase. I’m hosting this service off of the cable connection (what one might call a “best-effort” IP delivery service) I use for my home and business use, so I wanted to make sure I wasn’t exhausting my own capacity nor adversely affecting my neighbors.

It turns out my concerns were not well-founded. Aggregate bandwidth usage does not seem to scale directly with the number of users. The graph below looks very similar to the graphs I’ve been pulling for the last few weeks (eponine is the name of my switch):

Sure, the graph is a touch portlier than its older counterparts (of which I wasn’t wise enough to save, so take my word for it), but it does show my concerns to be without substance. Whew.

Unfortunately, other concerns have reared their ugly heads. My SLA with Comcast is something akin to a gentleman’s agreement that they’d be “sensitive” to any downtime. They’ve been good about that in the past; but I’m concerned that if my line were to flap for an extended period, my neighbor servers on EsperNet would not be able to handle the influx of users angrily forced upon them. While that wouldn’t rise to the level of disaster, the mere thought of it is, at least, disconcerting.

Oh well. One thing at a time.