A Load Balancing Scheme for Resilient Search in KAD Peer-to-Peer Networks

• Main Authors: Wu, Tai-Ting, 吳岱庭
• Other Authors: Wang, Kuo-Chen
• Format: Others
• Language: en_US
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/38423854649114875484

碩士 === 國立交通大學 === 網路工程研究所 === 97 === Kademlia (KAD) peer-to-peer (P2P) networks have been widely used in file sharing applications. However, these P2P networks suffer from the unbalanced publishing load problem. It causes a few peers handling large numbers of indexes. Those peers with high loads may become the bottlenecks of the network. Therefore, we propose a multiple hash method (called KAD-N) to balance peer loads in the KAD network. Note that N is the maximum hash times, determining by a cost-effectiveness factor. This method hashes the keyword of an object r times to produce a key for publishing objects, where r is a random number and 1 ≤ r ≤ N. Simulation results show that the distribution of indexes is more balanced using the proposed KAD-N method. We found out that N = 7 (KAD-7) is the optimal setting in our simulation environment. We used a standard deviation to evaluate the proposed load balancing method. Simulation results show that KAD-7 has the search hit rate close to 100% and the standard deviation is 44% less than that of the KAD (i.e., KAD-1), which means the proposed method is more load balanced than the KAD. However, KAD-7 has 7% extra traffic overhead. By increasing the search hit rate, KAD-N improves the search resilience of KAD networks with failed peers. Furthermore, the proposed KAD-N method can be easily extended to other DHT-based P2P networks.