In a peer-to-peer file-sharing system, a free-rider is a node which downloads files from its peers but does not share files to other nodes. Analyzing the free-riders’ impact on system throughputs is essential in exam...In a peer-to-peer file-sharing system, a free-rider is a node which downloads files from its peers but does not share files to other nodes. Analyzing the free-riders’ impact on system throughputs is essential in examining the performance of peer-to-peer file-sharing systems. We find that the free-riders’ impact largely depends on nodes behavior, including their online time and greed of downloading files. We extend an existing peer-to-peer system model and classify nodes according to their behavior. We focus on two peer-to-peer architectures: centralized indexing and distributed hash tables. We find that when the cooperators in a system are all greedy in downloading files, the system throughput has little room to increase while the cooperators throughput degrade badly with the increasing percent of greedy free-riders in the system. When all the cooperators are non-greedy with long average online time, the system throughput has much room to increase and the cooperators throughput degrade little with a high percent of greedy free-riders in the system. We also find that if a system can tolerate a high percent of greedy free-riders without suffering much throughput degradation, the system must contain some non-greedy cooperators that contribute great idle service capacity to the system.展开更多
基金the National High Technology Re-search and Development Program (863) of China(No. 2007AA01Z457)the Shanghai Science and Technology Development Funds (No. 07QA14033)
文摘In a peer-to-peer file-sharing system, a free-rider is a node which downloads files from its peers but does not share files to other nodes. Analyzing the free-riders’ impact on system throughputs is essential in examining the performance of peer-to-peer file-sharing systems. We find that the free-riders’ impact largely depends on nodes behavior, including their online time and greed of downloading files. We extend an existing peer-to-peer system model and classify nodes according to their behavior. We focus on two peer-to-peer architectures: centralized indexing and distributed hash tables. We find that when the cooperators in a system are all greedy in downloading files, the system throughput has little room to increase while the cooperators throughput degrade badly with the increasing percent of greedy free-riders in the system. When all the cooperators are non-greedy with long average online time, the system throughput has much room to increase and the cooperators throughput degrade little with a high percent of greedy free-riders in the system. We also find that if a system can tolerate a high percent of greedy free-riders without suffering much throughput degradation, the system must contain some non-greedy cooperators that contribute great idle service capacity to the system.
