Experimental I2PSnark build with more aggressive downloads

Issues and ideas about I2PSnark
Post Reply
zlatinb

Experimental I2PSnark build with more aggressive downloads

Post by zlatinb »

Hi,

I am working on a theory, outlined in ticket http://trac.i2p2.i2p/ticket/2280 that the default Snark request pipeline of 5 outstanding requests is too low. I've put together a build that raises that number to 7, which is the maximum possible without any server-side changes.

I invite you to test this build and am curious if it increases download speeds for you. It is available at Postman or clearnet. Please provide feedback here or on the Trac ticket above.

Note that this is a personal build by me, not endorsed by the I2P project. This experimental change may never make it into an official build.
BeaconLilt
Posts: 18
Joined: 05 Dec 2018 02:18

Re: Experimental I2PSnark build with more aggressive downloads

Post by BeaconLilt »

zlatinb wrote: 03 Dec 2018 05:08 I am working on a theory, outlined in ticket http://trac.i2p2.i2p/ticket/2280 that the default Snark request pipeline of 5 outstanding requests is too low. I've put together a build that raises that number to 7, which is the maximum possible without any server-side changes.
Some information that may pertain to this. I've made two identical batches of a set torrents, at various stages of download and some finished, in two folders. One I download only to normal I2PSnark. The other I use BiglyBY which is run under I2PD or I2P, not at the same time because they conflict. At other times I've run both I2PSnark and BiglyBT, under I2P, at the same time. They will both be using the same installed I2P system.

Here's some observations.

1.I2P(using I2PSnark) and I2PD(using BiglyBY) as the torrent download each alone alone seem to download at somewhat the same rates, Possibly with a slight edge to the I2PD.
2. Running I2PSnark then starting BiglyBT(it's torrents will lag as they have not been running at all and are in a different BiglyBT download folder), you can get massive download speeds between the two. Some random numbers but averaging what I've seen. With just I2P(alone) linked to random seeders I can get 800Bps, 1.2Kbps, from each seeder, maybe a little more but these values are common. With a I2PSnark that has run for a good while to build up completion to it's download folder then turning on BiglyBT after the DHT links up you easily get 40KBps, sometimes 90KBps. BIG DIFFERENCE. This should not be. Why isn't it doing this all the time???? BiglyBT is the one catching up. Also it seems that other torrents speed up somewhat in I2PSnark at the same time(I've only seen this once. It may have happened other times but it caught my attention because it seemed the normal rates from unknown seeders leapt up to a much better, 100-200% better rates. I could have missed it at other times so I can't say this happens all the time but once it did.)
3. The counter also is the same. If I stop several torrents in I2PSnark while letting them run in BiglyBT and then starting I2PSnark I will get high download speeds to I2PSnark from BiglyBT. Why???

Now this leads to the idea that I have low seeding ratio, accounting form the speed up of not only the download to BiglyBT but a speed up from unknown seeders to my I2PSnark WHILE I have high speed seeding TO BiglyBT. But I leave seeds on a long time. The ratio shouldn;t be that bad. Mostly when I stop them no one is asking for them. I might also add in this same forum others seem to have the same problem who are seeding a LOT but don't seem to be able to seed at any great rate as they have done before. So could there be some sort of conflict in the ratings system? So that I2PSnark to I2PSnark is too harshly rating each other so neither will seed at the rate that they can do so???? BiglyBT used on the same set of files and running less actually gets better downloads from a local I2PSnark. Even though it's local it still has to go through other servers. It, I assume, doesn't know it's local. There's also the fact that halting some of I2PSnarks torrents then seeding them with BiglyBT results in high download speeds. Why wouldn't it be affected the same as some random I2PSnark seeder out there???

Maybe what you said about the bandwidth limits being so low is the whole problem and before, somehow, the limits were ignored. When the timer problem was fixed maybe it kicked in a more severe bandwidth limit that before was skewed because of timer problems. I have my output bandwidth up high on both I2PSnark and BiglyBT.

Of course this would not apply if the poster that had set his system up to feed many torrents and has delays feeding them has his limits set high also.

I'm not saying what you are doing is not good by any means, or even implying it, but I'm not so sure that that's THE reason for the slowing of torrents as I do just fine with two systems feeding each other on the same machine which is a fairly good test of what I2PSnark CAN do and it's very good and fully adequate, but it's only one system and I can't know what others are doing.

It's very mystifying. Sorry for posting such random stuff and being so wordy but...it's seems answers to odd things like this sometimes lie in in obscure places so I tried to comment on what I've seen that seems contrary.
BeaconLilt
Posts: 18
Joined: 05 Dec 2018 02:18

Re: Experimental I2PSnark build with more aggressive downloads

Post by BeaconLilt »

I downloaded this and unpacked the ,jar. I don't see where I can save a portion of my install and over write it with this one. I appreciate all the work you've done on this but I don't want to overwrite what I have installed now. If I could save what I have to a new folder or even better what if you made it a portable install? All the files needed in one folder. If you had a portable install that wouldn't change anything else I bet it would be tested much more. I can't be the only one that's tweaked their I2P install. Think of all the stuff you've changed in yours and ask if you want to write all that stuff down and redo it over.
Post Reply