以滿足度為服務水準之Overlay Network 在 P2P 檔案分享最佳化

Abstract

Overlay Network問題是目前P2P系統評量其效率的主要因素。使用者藉由其所屬Overlay Network的連結，以提升在P2P使用者之間的工作效率。在此篇論文中，我們最佳化Overlay Network應用於P2P 檔案分享。我們設計了以檔案下戴滿足度作為整體Overlay Network的服務水準，追求整體P2P系統頻寬分配的最大化。 本研究中，我們使用粒子群演算法來求解．並且設計另外四種經驗法則來進行比較。我們透過不同的服務水準數值來評估本研究之效益。由實驗結果顯示，每一個經驗法則在不同的系統要求中，會產生不同的優劣情形。此外，這四種經驗法則均不能同時滿足不同環境的要求。而本研究所設計的粒子群演算法能在不同環境下獲得可行解，同時執行時間迅速且不錯的目標值。

The Overlay Network Problem is a key factor in performance evaluating of P2P. A peer can enhance the efficiency in P2P by using the connections of overlay networks. In this paper, we optimize overlay networks for file-sharing P2P systems, by maximizing the bandwidth allocation of P2P subject to the Service Level requirement based the completeness of the content fragments. In this study, we apply PSO (Particle Swarm Optimization) to drive near optimal solutions, and we also develop four heuristics to compare the quality of solution. We consider different service levels based the completeness of the content fragments to evaluate this problem. The results of experiments reveal that each heuristic has its advantages under different system requirements and none of the heuristics can satisfy different requirements at the same time. The PSO algorithm can obtain feasible solutions with impressive quality in a short time.