Web seeds are slow or do not connect in a hybrid torrent

When downloading a V2 torrent from the Internet Archive, it seems that webseeds are never connected to and the torrent never starts, or stalls after a few moments.

I suspect most torrents from the Internet Archive that are recently derived will have this issue.

this.TorrentClient = new ClientEngine();

var torrent = await this.TorrentClient.AddAsync(
                await Torrent.LoadAsync(torrentStream), 
                Path.Combine(Environment.CurrentDirectory, $"download"),
                new TorrentSettingsBuilder()
                {
                    CreateContainingDirectory = false,
                    
                    // encourage webseed use
                    WebSeedDelay = TimeSpan.Zero,
                    WebSeedSpeedTrigger = int.MaxValue,
                }.ToSettings());
await torrent.StartAsync();
await Task.Run(async () =>
            {
                while (torrent.State != TorrentState.Stopped && torrent.State != TorrentState.Paused && torrent.State != TorrentState.Seeding)
                {
                    if (cancel.IsCancellationRequested)
                    {
                        await torrent.StopAsync();
                        cancel.ThrowIfCancellationRequested();
                        break;
                    }
                    await Task.Delay(1000);
                    this.SpeedChangedEvent?.Invoke(this, torrent.Monitor.DownloadSpeed);
                    this.ProgressChangedEvent?.Invoke(this, torrent.PartialProgress);
                }

                var assetFile = torrent.Files.Single(f => f.Path.StartsWith(asset));
                return File.OpenRead(assetFile.FullPath);
            });

[alanmcgovern]

I updated https://github.com/alanmcgovern/monotorrent/pull/540 to make these timeouts configurable. I also increased the default values as per original PR.

Thanks for the report and for testing the tweaks! Hopefully that’s the only webseed issue you see :p

[alanmcgovern]

That’s what I experienced too.

The way to improve that is to request more data per http invocation. I’d have to double check, but it looks like the data was being requested in chunks of 4MB. There’s a latency of about 10-15 seconds for the server to accept the http request, which gives an effective throughput of about 0.5MB/sec. There are two http seeds so that gives a total throughput of nearly 1MB/sec.

Requesting I’m higher chunk sizes would be less beneficial if there were other peers sharing data. It would also be less beneficial if the server had “normal” latency when serving the response data.

Do you see any reasonable options to making this better, or is this the status quo after the timeout changr good enough that the webseed fallback is good enough for your needs?