基於基因演算法在點對點即時串流系統下之動態最佳化

Abstract

在點對點網路架構下節點可以從多個來源接收視訊片段，節點如何選擇具有足夠能力的父節點來滿足自身的需求是一項重要問題。過去的演算法大多使用單一的評定方法來挑選父節點群，而這篇論文提出一個可以同時衡量多個因素，並且針對節點的目標需求與評估當時的網路環境，配合基因演算法實施動態最佳化策略。模擬的結果顯示在符 合使用者的期望下，其他次要目標也能夠維持一定水準。

In Peer-to-Peer network structure, a peer can receive video fragments from several source nodes. It is an important issue that how a peer pick up group of parents to fulfill its demand. Many algorithms in the literature adopt some monotonic criterion to choose parents before. In this paper, we propose an algorithm that integrates genetic algorithm for dynamic optimization with multiple criteria and multiple objectives. The simulations show the proposed algorithm can satisfy the demand/objectives defined by any individual peer.