Buffer Overloaded Reduction Mechanism in a Gossip-based Broadcast System

• Main Authors: Bo-Chung Chiou, 丘博仲
• Other Authors: Tzong-An Su
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/76783600686577443753

碩士 === 逢甲大學 === 資訊工程學系 === 103 === Gossip-based Broadcast Protocol is also called Epidemic Algorithm, much like virus-spreading in the real-life, message disseminated by a gossip protocol system spreads effectively and reliably, with high arrival rate and system scalability. However, most research relies on assumptions that gossip protocol system is supposed to operate on expected environment. For instance, assuming that all of nodes have sufficient buffering resources and gossip never exceeds these buffering capacity limits, or the system runtime at different processes have the same rate, the processes gossip in real-time and don’t have any time delay. But the buffering resources at each node are limited in practical situations, it might impact the overall performance of the gossip-based broadcast system. Therefore, we propose an effective algorithm to estimate the buffering resource availability of each node to adjust the rate of messages emission accordingly, and reduce the number of messages being dropped in the dissemination because of nodes do not have enough resources, and prevent unnecessary overhead of multiple deliveries of the same message. In the environments of limited resource, this research makes the Gossip-based broadcast system still have high performance of propagation.