Distribution on trackers, where it is difficult to hand out, or how to defeat a sideboxer

As you know, some torrent trackers have a so-called rating. 5 years ago, when unlimited Internet had just appeared in most cities of Russia, and its speeds were very low, the rating on some trackers was very difficult to maintain. But what is there to support, even at least something to hand out is difficult. And then, in search of rare music, I hit just such a tracker.

The channel I had was 512/512 kilobits / s (symmetric). Almost immediately, I discovered that on other people's distributions with such a channel, little can be distributed. More precisely, practically nothing. Of course, you could start creating your distributions, but I decided to find out if it was possible to do something with strangers. It turned out that it is very possible.

Although many years have passed, I still see reports of rating difficulties, and I personally know several people who still do not dare to download anything from this tracker, fearing that it will not work out. Therefore, I think this information will be useful. The above screenshots and graphs were made not 5 years ago, but now, to schematically show the events that took place then.

Naturally, with a slow channel, you can not download several hands at the same time, only one. But even if the distribution is not conducted by the sidebox , but, for example, a person with a channel of “only” 1-2 megabits, the following picture turns out: those who download very quickly give me what they have already downloaded from the sid, this stream easily fills my entire channel, and I’m sitting gives nothing away. Accordingly, I do not distribute anything.



It can be seen that the segments that everyone already has are swinging (in the screenshot availability = 20).

Then I tried to find slow hands. In order not to connect to everyone in a row, and then quickly disconnect when it becomes clear that there is nothing to catch, I found this way: we look at fresh distributions, and if we see that there is one seed and several leechers, moreover, the distribution has been going on for 1-2 hours , but they haven’t downloaded it yet, so the seed is giving out slowly, and here you can win something. Usually this happens with fairly large distributions, flasks , and preferably multi-disk - the larger the volume, the more room for maneuver (more on this later).

It turned out that even if the seed has an extremely low speed (for example, like mine, 60KB / s), then there is nothing to catch on the distribution anyway. Because by the time I connect to it, the leechers have already downloaded something, they immediately start distributing it to me, and again the same picture turns out. Even if I manage to “catch up” them (and this is not easy) and the seed begins to give me something, the leechers very soon have new pieces, and they actively “push” them into me, immediately clogging the channel, and everything repeats again. Thus, in order to distribute something, it is necessary that the speed of distribution is 2 times less than mine, that is, about 30kb / s. And this happens on this tracker is extremely rare.

However, there is a way out. You can download only part of the songs from the distribution, for example, only one. This was not forbidden by the rules of this particular tracker. And this is what the process looks like. We find the distribution (see above), let's say this is a 500M flask, out of 10 songs, it has been downloaded for 1 hour and downloaded by 30% - a typical picture. I'm starting to download one song. At first I “catch up” with other leechers, the whole flow comes from them, distribution is zero. Since I download only one song, I quickly download everything they have from it (a third of a 50M song is 17M, it can be downloaded in 5 minutes). After this, a lull usually follows, since the leechers have nothing more to give to me, and the seed "does not notice me" so far. This usually lasts a few minutes, but sometimes longer. Sometimes there is a seed that does not “see” me at all, apparently completely carried away by other leaches, but this is rare.

So, now the seed is starting to give me a little bit of the first piece. This happens slowly, because, for example, if its upload speed is 80KB / s, and only 6 leechers, then everyone gets 13KB / s. Sooner or later, a piece comes, and (lo and behold!) All the other 6 leechers (usually seedboxes) attack him. And they attack at the same time, in the ideal case, taking each piece as a whole, that is - I downloaded 1MB from the sid, distributed 6MB. In practice, this is not always the case, they still have different speeds, and sometimes some manage to merge my piece earlier, and distribute it to others. And also do not forget that although I download only 1 song out of 10, others also download it. And when they download a piece from “my” song, they immediately give it to me, which often leads the sid “to confusion”, and for a while it stops giving it to me again.

The traffic graph looks something like this:



Here, the incoming traffic is shown in red: narrow impulses are visible - segments coming periodically from other peers, and slowly but constantly coming data from the sid. The distribution is shown in green: the segments periodically leaving simultaneously to all peers merge into wide pulses of outgoing traffic.

In general, in the end, after 2 hours of distribution, it comes to an end, everyone downloads the album completely, and I download one song, having 50MB down, and from 50 to 200MB up, on average, the rating on each distribution is about 2. At the same time it must be borne in mind that the speed of distribution can also change: both increase and decrease, while it is desirable to respond adequately. If we see that our only song is coming to an end, and the rest have not yet downloaded everything, you can add a second, third, etc. However, one should not be forgotten, it is only necessary to slightly exceed a certain invisible threshold (“order” too many songs), as pieces coming from other leechers clog everything and the return drops sharply to zero.

The graph above is valid for the case when the speed of the sid channel is much lower than mine. If his speed is higher, or I “order” more songs for the jump, there will be more red pulses, and if they begin to block my entire channel, the green recoil pulses will practically disappear:



Then we look for the next distribution and everything repeats again. Stubborn experiments allowed me to learn how to accurately choose not only the number of songs, but which songs to download based on their size in order to fill the outgoing channel completely. Of course, not at any distribution you can achieve this, but having enough experience and knowing all the parameters (speed of distribution, its size, size of pieces, speed and “behavior” of each feast), you can achieve this sensitive balance when the next piece coming from the sid is right there is distributed to the evil sideboxes, and while it is being distributed to them, the next one comes from the sid, and so on and so on, and by the end of the distribution I just have time to get my 1-2 songs of the right size, downloading a minimum of megabytes in the end and distributing this is the maximum that could be squeezed. But for this, in fact, you need not to leave the screen, all the time,

Of course, there are still different nuances, for example, feasts with the libtorrent client (which tend to download distribution from left to right), and you need to select songs that they have not “covered” yet, and better with the expectation that they will reach them later.



Here you can clearly see the continuous area, which is downloaded by the sidebox with libtorrent, and the rest of the torrent is downloaded by other, "normal" clients in random order. And sometimes other tricky people come across who also download only one song. In this case, you should try to choose not what they download. It should be remembered that sorting songs inside a torrent can be either sequential or random (this is determined by the distribution author when creating the torrent). So, for example, from the above figure we can conclude that the seedbox most likely downloads song number 9 or 10. So you need to start with song 8. If the location of the files is random, then this can only be determined empirically.

And from time to time among distributions “the best distributions” come across - where there are no seedboxes. Since I always downloaded only slow distributions, sooner or later I got to such ones. And then the real holiday begins. At this distribution, while it is still going on, you can download many songs, and sometimes the entire album. And it even happens that in general everything goes “in plus", that is, the seed distributes only to me, and I calmly pass it all on to the other lecturers, even slower than me. Well, the subsequent distribution in this case is also very productive, unlike distributions with seedboxes, on which, as you know, only they give out.

For about 10 days, after examining the intricacies of peer-to-peer exchanges, I distributed 20 gigabytes, and at the same time I received 1-2 songs from albums of various styles that I probably would never have known about, and then calmly downloaded those that I liked . It is not necessary, as described above, to try to squeeze the maximum out of each torrent, it is enough to at least distribute a little. And then, in combination with the usual tips for any trackers (to stay in the distribution, create your own, etc.) over time, you can achieve good results.

ps I remind you that this is an article about trackers, where it’s hard to pass out, where the usual methods of maintaining the rating do not help.