A dynamic load balancing strategy with the push and pull approaches in DHT networks

The dynamic load imbalance problem, probably caused by the heavy-tailed distribution of file requests, negatively impacts on the distributed hash table (DHT) networks’ availability. The existing solutions mainly employed the local load information to design the load balancing strategies, which often need to calculate the peers’ loads and execute the balancing procedures periodically, and thus their effectiveness could not be guaranteed and network bandwidth is wasted. To address this problem, we first describe the mechanisms for managing the download volume and the upload volume of each peer, as well as the information of the heavily loaded nodes and the lightly loaded nodes classified by double thresholds, and then we present a novel load balancing strategy which transfers the loads from the heavily loaded nodes to the lightly loaded nodes with the push and pull approaches. The simulation results show that our scheme is effective and efficient in handling the load imbalance problem in DHT networks.

Graphical abstractFigure optionsDownload full-size imageDownload as PowerPoint slideHighlights► We propose a strategy for collecting download volume and upload volume of each node. ► We present a mechanism for managing the information of heavy nodes and light nodes. ► Push and pull approaches are employed to effectively and efficiently balance the load.