A long time ago I was using a 100Mhz machine to serve as a networked torrent slave; it’s a fantastic use for an out-of-date computer and one that requires almost nothing in resources. The example I built had only 16Mb of memory and almost nothing else, and it worked fine … more or less.
I’ve experimented a little more with rtorrent since then — in particular watching the system load on a 120Mhz machine with 80Mb of memory. Here’s a little picture of rtorrent with htop running alongside it in Ubuntu 9.10, so you can get a mental picture of what’s happening.
I have about 12 torrents running at once there, some seeding and some downloading. The idea was like the one I had a few months ago, just to dump as much as possible on rtorrent and see what kind of space it takes up.
Keeping in mind that my normal desktop “suite” is also running , the greatest strain on the processor seems to come when bandwidth is high. Just dumping torrents on it didn’t make it flicker at all, until the network transactions began to climb.
At the same time, memory load there is higher — much higher, in some ways — than what is normal for me, and that seems to come about when the torrents are added. Memory demands increased sharply as I added torrents, but fell away just as quickly when I removed them one by one.
The only reason this is worth mentioning is because some machines — particularly early Pentiums, like the one I was using — have a fused set of 16Mb or so built into the machine, and are expandable to a small degree beyond that. If you can’t get the right chips (SIMMs or S0DIMMs whatever they were called in your generation) and you’re forced to rely on a very slim memory profile, you’ll have to likewise manage the number of torrents you handle. Too many torrents will push the machine beyond its memory limits.
At the same time, a few popular torrents might push the processor to its maximum continually, and that too can be a detriment. The obvious way to prevent that is to throttle the upload and download controls, so there’s less chance of overwhelming the machine and causing a system failure.