以動態排程指數法提升虛擬化網頁伺服器叢集效能

Virtualized Web Server Cluster Scheduling Scheme Adjusted by Dynamic Indexes to Improve the Performance

陳智民(Chih-Min Chen)

指導教授 ： 詹昭文

國立臺中科技大學/資訊與流通學院/資訊工程系碩士班/碩士(2012年)

https://doi.org/10.6826/NUTC.2012.00091

摘要

多核心處理器的普及化，帶動了虛擬化技術快速發展。透過虛擬化達成伺服器合併，可以降低硬體的購置成本，提高伺服器運算資源的使用率，簡化基礎架構管理的複雜度與減少能源的耗用量等效益。

　　伺服器虛擬化的應用範圍相當廣泛，包含網頁伺服器、郵件伺服器、DNS伺服器、FTP伺服器及資料庫伺服器等。透過虛擬機器(Virtual Machine，VM)來建構網頁伺服器叢集，使得硬體資源的運用將更具彈性。

　　本文提出一個動態排程指數法(Dynamic Scheduling Indexes Scheme)，此方法係藉由即時監控虛擬化網頁伺服器叢集內可用伺服器的負載資訊，包含處理器、記憶體及網路頻寬的負載值，接著將負載值計算成負載排程器分配的排程指數，負載排程器依據排程指數來分配用戶端的連線請求，之後固定每60秒進行動態調整分配的排程指數。

　　實驗結果顯示，相較於請求計數排程法(Request Counting Scheduling)、權重流量計數排程法(Weighted Traffic Counting Scheduling)和擱置請求計數排程法(Pending Request Counting Scheduling)，我們的動態排程指數法有較為良好的表現，的確能夠提升虛擬化網頁伺服器叢集效能及縮短網頁連線請求的回應時間。

關鍵字

虛擬機器 ； 動態排程指數法 ； 網頁伺服器叢集

並列摘要

The popularity of multi-core processors has led to the rapid development of virtualization technology. The implementation of server consolidation through virtualization can result in a number of benefits, such as reduced hardware acquisition costs, improved applications of the computing resources on servers, simplified infrastructure management, and reduced energy consumption.

　　Server virtualization can be applied in a variety of server types, including Web servers, mail servers, DNS servers, FTP servers, and database servers. By constructing a Web server cluster on virtual machines, hardware resources can be employed with more flexibility.

　　This thesis presents a Dynamic Scheduling Indexes Scheme that continuously monitors load information concerning available servers in a virtual Web server cluster, including information about the load values of CPU, memory, and network bandwidth, and then converts these load values into scheduling indexes that can be allocated by the scheduler server. Based on scheduling indexes, the scheduler server forwards client requests to Web servers, and then dynamically adjusts allocated scheduling indexes every 60 seconds.

　　Experimental results indicate that, for performance, our Dynamic Scheduling Indexes Scheme is superior to Request Counting Scheduling, Weighted Traffic Counting Scheduling, and Pending Request Counting Scheduling. The present scheme has been verified to improve the performance of a virtualized Web server cluster and to reduce the response time of client requests.

並列關鍵字

Virtual machine ； Dynamic Scheduling Indexes Scheme ； Web server cluster

參考文獻

[2] V. Cardellini, M. Colajanni and P. S. Yu, Dynamic Load Balancing on Web-server Systems, IEEE Internet Computing, pp. 28-39, 1999.

[5] Y.M. Teo and R. Ayani, Comparison of Load Balancing Strategies on Cluster-based Web Servers, Transactions of the Society for Modeling and Simulation, pp. 185-195, 2001.

lb_algorithms.pdf, 2012.

[8] D. Mosberger and T. Jin, httperf: A tool for measuring web server performance, In Proceedings of the Workshop on Internet Server Performance, pp. 59-67, 1998.

[18] P. Wang, Approaching The Cloud Computing, Posts and Telecommunications Press, 2009.