查詢結果分析
來源資料
相關文獻
- 設置影音串流分散式系統之最佳化演算法
- 配電饋線三相不平衡分析及改善策略之研究
- Design Optimizations of Turbomachines Using Evolutionary Algorithm
- 應用啟發式演算法於股票選擇權避險策略最佳化之研究
- 應用修正式蜂群最佳化演算法求解撥召問題--以復康巴士問題為例
- 混合複數類神經模糊與自動回歸差分平均移動方法之智慧型時間序列預測模型
- Ant Colony Optimization for Railway Driver Crew Scheduling: from Modeling to Implementation
- A Dynamic Local and Global Conjoint Particle Swarm Optimization Algorithm
- 粒子群最佳化應用於雲端服務提供之解決方案
- Measurement and Analysis Techniques for Designing Microwave Absorbers
頁籤選單縮合
題 名 | 設置影音串流分散式系統之最佳化演算法=Optimization Algorithm of Installing Streaming Media Distributed System |
---|---|
作 者 | 顧淇元; 王姿婷; 張治雄; 申曉玉; 李錦烈; | 書刊名 | 電信研究 |
卷 期 | 33:3 2003.06[民92.06] |
頁 次 | 頁413-427 |
分類號 | 448.889 |
關鍵詞 | 最佳化演算法; 影音串流分散式系統; Optimization algorithm; Streaming media distribute system; |
語 文 | 中文(Chinese) |
中文摘要 | 為配合ADSL寬頻網路的興起,提供HiNet用戶更豐富的影音節目內容起見,建置網路多媒體視訊應用服務,拓展網路加值服務。採用影音串流技術,透過IP多點廣播來進行視訊廣播,提供隨選視訊、即時節目轉播、教育訓練課程、定時電視節目播放等新式服務遂成為目前最重要之課題。惟建置影音串流服務系統時需考量建置成本及影音品質以取得雙贏為原則,未來若採用集中式架構,雖有方便管理之優點但各地用戶點播節目時卻需透過網際網路至服務提供點擷取影音資料,不僅增加幹線之負荷,而且使得影音品質變得無法預測。為減輕長途幹線Traffic的負荷量,並改善影音品質起見,推出分散式架構,在接近用戶端之各POP點之數據機房設置Cache Server與Distribute Storage,提供影音串流服務,以節省長途幹線成本負擔。 分散式架構雖可節省長途幹線成本之負擔,但需在接近用戶端之各POP點設置Cache Server與Distribute Storage等設備,卻也增加影音串流設備之成本。兩者之間如何取捨以取得最大的利基乃為本文之重點,本文探討之成本涵蓋長途電路設備如線路、機房、電力、土地、電力設備以及regenerator等、影音串流系統設備成本如Central Farm之Manager Server、ICP Creating Server、Web Server、AAA Server、Live Server、VOD Server、Media Direactor、Database Server及Central Storage以及Distribute Nodes之Cache Server、Live Relay、Media Director及Distribute Storage等元件成本,由於影音串流系統元件成本係採分散式架構之產品作為分析之依據,不同廠牌之產品只要將相關之參數如Cache Server、Media Director及Distribute Storage與Relay Server等加以調整後即可適用。 本文所提之演算法係參酌各POP點與Central Farm間之長途電路設備、影音串流設備以及未來之Concurrent On Demand Traffic等因素,計算出最佳建置Distribute Nodes之數量、位置及各POP點管轄權歸屬之方法。依此方式計算,不僅可估算出目前所需建置Distribute Nodes之數量、位置及管轄權歸屬,未來擴充時適合之建置位址及管轄權移轉等亦均適用,可做為中華電信建置影音串流系統之參考。 |
英文摘要 | In order to provide rich streaming video contents for the HiNet ADSL clients, a streaming video system shall be installed in the HiNet to accrue to the value added service network. This streaming video system streams live multicast, Video On-Demand, e-learning, scheduled programs to the Internet clients through IP network. However, it is important to consider the cost as well as QoS (Quality of Service) of the video programs. Central Farm Co-locates all media servers, storages, database servers and manager servers in one site, it is convenient to install, manage and monitor the system behaviors, but unfortunately, the backbone bandwidth was exhausted while thousands clients concurrently watch video programs, the Quality of the video will be degraded by the traffic jam. Distribute Nodes could improve the bandwidth starvation problem but cost more on installing the Cache Servers and Distribute Nodes. This subject proposes a Distribute system to improve the video quality as well as save transmission cost in the backbone as following chapters. Cache Servers and Distribute Storages in the said Distribute system are installed in the Distribute Nodes to pump video streams to the requested clients. More Distribute Nodes stream higher quality video to the dedicated clients but cost more. It is hard to deal with video quality and investment cost, however, an algorithm has been proposed by our laboratory to find a good location to install Distribute Nodes in order to save investment cost, provide good quality video and reduce the backbone traffic. The said investment cost includes of transmission cost, such as long distance circuit, central office, electricity, land space, power generator, optical regenerator and so forth, and streaming media appliance cost including Manager Server, ICP Creating Server, Web Server, Live Server, VOD Server, Media Server, Database Server and Central Storage for the Central Farm; meanwhile, the Cache Server, Live Relay, Media Director and Distribute Storage for Distribute Nodes. The Specification of the streaming pumping servers are for general purpose, it could be reused at any case. The referencing factors of this Algorithm include of long distance transmission cost (from POP to Central Farm), streaming media appliance cost and the forecasting future Concurrent On-Demand traffic. Using this Algorithm, it is convenient to evaluate the best location of installing a Distribute Nodes as well as the amount of Distribute nodes and its related POP. It is useful to relocate any new Distribute Node among the existed Distribute Nodes topology as well. |
本系統中英文摘要資訊取自各篇刊載內容。