In this paper, we propose a source selection algorithm for multi-source downloading in peer-to-peer networks. It is unreliable that a receiver downloads a file from one source, and the downloading speed will be slow because of the characteristic of the peer-to-peer networks. Therefore, it can be an efficient way that a receiver downloads from the multiple sources which have the same file at the same time. If there are many sources, we have to solve the problem to select some of them.
Downloading from all sources can be inefficient, because they share the transmission routes. If multi-sources share the path, transmission speed cannot exceed the bottleneck bandwidth of the shared one. Therefore, selecting all sources does not serve maximum downloading speed. The goal of source selection is to reduce packet drop on the shared path and the network overhead.
We define the average downloading distance in multi-source downloading. Long average downloading distance means that more routers in the network should forward the packets, so it increases network overhead. If there are enough sources which have same file, it is possible to decrease the average downloading distance by source selection which preserves downloading speed.
Our experimental results show that the throughput of multi-source downloading increases about 20% compared to downloading without source selection. Proposed source-selection algorithm shortens the average downloading distance and avoids packet collision on shared path.
