Improving DTNs Performance by Reduction of Bundles Redundancy using Clustering Algorithm

摘要

This article presents complex clustering algorithm for Delay Tolerant Networks (DTNs) including neighborhood discovery, cluster creation, and data distribution within cluster. The general idea is to reduce the amount of messages being sent and of buffer utilization, by taking advantage of the nodes tendency to create groups and to share similar mobility patterns among each other. The main purpose of the algorithm is to improve the network performance without major changes in communication schemes between nodes. Almost no extra message type is added. Most extra features are available thanks to adding small extra fields into transmitted packages. Neighborhood discovery is being realized passively by listening to other nodes messages. The proposed algorithms allow to reduce both the bandwidth occupation, as well as the problems related to the media access. Furthermore, it can increase message delivery probability thanks to intelligent package distribution inside created cluster. Simulations were carried out to evaluate the effectiveness of the proposed solution in terms of package delivery probability, mean buffer occupancy and mean hop number to delivery the message. Results of simulation show that this solution is not necessary or recommended for small-scale networks with few nodes using clustering algorithms. However, with increasing number of nodes and messages, the performance of non-clustered DTNs drops significantly while clustered network works efficiently.