Wednesday, July 18, 2012

Multi-Tracker Ignorance (Or Why You Don’t Need More Than One Tracker)

There are people out there who run torrents with half a dozen trackers on them. There are some that will even add 100+. The reason, they claim, is to ‘find peers’, but in reality they’re just causing a massive problem for trackers, and themselves.

When people add trackers to torrents, they do so in the expectation that they will find more peers, and that it will help them. In fact, the image in their mind is pretty similar to this image.
Image based on one by trackerbt. All rights theirs.
It's not actually the case. It can be hard to visualize though, so perhaps a metaphor would work better.

Let’s move away from abstracts of coding and peers, and move into the real world. Instead of peers and trackers, let’s have people and meeting places. With a real-world example, you can probably visualise things better.

Let’s say you’re in to rock climbing. You go to the local rock climbing club, and you meet the people there. This club is like the tracker. It’s where you go to find people interested in the same thing (like going to a tracker to find peers for the same bittorrent hash).

All fine and dandy. However, you’re not satisfied with those people; you want to meet more and different people to go rock climbing with there and then. Do you turn up at a random cliff face/wall? No, because the likelihood of there being people there you’re interested in is low. Do you go to the pub? While, yes, there are people there, it’s unlikely they’re interested in going rock climbing, even if you sit at a table and have a sign saying “who wants to go rock climbing right now”. You’ve added two places (trackers) and found no new people (peers) interested in your activity (or torrent).

Now, you COULD go to your club and say “I’m going to the pub to find people who want to go rock climbing”. Some people from the club might go with you. However, still no-one at the pub is interested in going. There may be a bigger group of you wanting to rock climb at the pub, but they’re the same people from the club, you’ve gained nothing, and wasted time and effort.

Why is there no-one at the pub who wants to go rock climbing there and then? The answer is that people who want to rock climb go to a rock climbing club, they don’t go to a pub.  It’s not a known or expected point of contact for people who want to go rock climbing. You have, however, used up ‘resources’ (bandwidth and connections for a tracker, a table in our metaphor) that may have inconvenienced someone who would have made productive use of it.

This is the problem with adding random trackers. Unless people know to go to them, they don’t. Instead if you want to find more interested people, you use a general finding approach, such as a magazine, or the internet. You cast a wide net, and leave a way to contact you, rather than a specific point of contact. This is what DHT does.

So next time you think about adding multiple trackers, stop, and think. Don’t do it, just add DHT, it’s better for everyone. Don’t think adding trackers helps, because it doesn’t.

As a side note, how many clients count peers across trackers is an important thing to understand why this phenomenon persists. If there are 50 of you in total at the club, and then 4 come with you to the pub, it’s still the same 50 people. However, if you count it as 50 at the club, and 5 at the pub, you’ll come up with 55 total. It might seem you’ve added 5, but you’ve just counted some twice. This is how many clients count total stats, and also goes for DHT and PEX values, which can make overall totals very unreliable (and why adding 130 trackers would make it seem like you’ve found another 100+ peers – Congratulations, you've found yourself!). It's only very rarely do you actually gain peers, despite the number you see.

1 comment:

  1. rubbish. your premises are wrong to begin with....