what-when-how
In Depth Tutorials and Information
he mean in this experiment, shown in Figure 6.8, progresses linearly towards
a higher value, finishing very close to 1. Figure 6.9 shows the rapid progression of
the block variation towards a very narrow distribution of pieces around the mean
value once the irst copy of the ile is present in the network (at 9 hours). he large
variation in the piece population in the first 6 hours is expected, as the piece popu-
lation goes from an initial stage of very narrow (all pieces have no copies), to a split
population (some pieces have no copies), and finally to a normal distribution of
narrow size soon after all pieces enter the system. he block variation again reaches
a value close to 10
-4
by the end of the experiment.
6.3 NeighborSelectionandNetworkTopologies:
TheIntra-TorrentLevelExamination
We proceed with the measurement of BitTorrent's neighbor selection algorithms
along with the topology of the networks formed. In particular, the resilience to
failing and misbehaving nodes suggests that the network may be
scale-free
, and the
efficiency of information distribution suggests that the network may be clustered or
even
small-world
. Neither of these properties has been quantitatively measured in
BitTorrent, and never beyond the early stages of torrents. Since BitTorrent networks
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
2
4
6
8
10
12
Time (hours)
Figure6.8
ThemeanofthepiecepopulationinthePlanetLab-1torrent(error
barsare95%conidenceintervals).
