On the performance of download strategies in a P2P like network

We investigate the relationship between average download time, system utilization and the level of competition among downloaders in a P2P like network. We show that when the system is nearly 100% utilized, the average download time increases with the level of competition. When the system is under-utilized, using parallel connections can help increase the utilization and hence reduce the average download time. However, contrary to common belief, we show that parallel downloading does not always help; it only helps if the system is under-utilized. We also show that even when the global metrics, such as competition factor and system utilization, are identical from a macroscopic point of view, the microscopic performance for each downloader can be very different depending on different downloading strategies.