點對點系統中群集大小與效能的研究

Abstract

點對點網路技術是網際網路上最重要的應用模型之一。在過去十年間，提出了許多不同的結構化及非結構化的點對點模型，然而不論是結構或非結構的模型，各有其優勢及劣勢之處。我們提出了一個混合式的模型，希望透過節點群集，在兩種模型間找到一個平衡點。既然節點群集的數量會影響系統效能，我們也分析了群集數量和維持網路結構的成本之間的關係。其中維持網路結構的成本是由維護時產生的訊息來計量。根據實驗結果，使用理論分析得到的群集數量非常接近實際的最佳值。而且混合式的模型可以比Chord使用更少的訊息，來維持網路結構。

P2P network is the one of the most important application models in Internet. Numerous structured and unstructured P2P models have been proposed in the last ten years, and both models do have distinctive advantages and disadvantages. We proposed a hybrid model to strike a balance between these two models using peer grouping. Since the size of peer groups is essential to performance, we analyze the effect of group size on the maintenance cost, which is measured in terms of the number of maintenance messages. Experimental results suggest that the group size obtained from our theoretical analysis is very close to the actual best group size obtained from simulation, and the hybrid model uses less messages in maintaining a P2P network than Chord does.