Blog

Posted 2006/03/31

Tycho File Swarm

Work continues on the implementation and I have started to look at how other people have tested this kind of system.

To assess how well my algorithms work (such as the one for selecting which chunks to download) and to test for concurrency problems I am running multiple peers and seeds. The important information in this table is much the times change in different configurations.


Configuration | Secs to download

1 seed, 1 peer | 15.21
1 seed, 2 peers | 17.83
1 seed, 3 peers | 20.95
1 seed, 1 peer | 15.21
2 seeds, 1 peer | 12.67
3 seeds, 1 peer | 12.34
4 seeds, 1 peer | 12.98

After I have tried out larger numbers of peers and clients we can start looking at the effect on downloading performance with different chunk and file sizes (number of chunks). This testing will form the basis of the benchmarking code.

During this process I found a bug in the Tycho library itself, it was a logic bug in the IRC connector which meant single line query replies would not be sent back to the client.