雲端主機監控系統負載平衡架構之效能分析

Performance Analysis of Cloud Host Monitoring System on Load Balancing Frameworks

陳竹正(Jwu-Jenq Chen) ； 張文貴(Wen-Kui Chang) ； 吳永基(Yung-Gi Wu) ； 俞怡中(Yi-Chung Yu)

《長榮大學學報》 15卷1期 (2011/06) Pp. 121-138

https://doi.org/10.30115/JCJCU.201106.0007

摘要

網路服務仰賴其伺服器，如何有效分散伺服器的負載，使其穩定的提供服務，這是雲端技術極為重要的課題之一。但為了達到網路尖峰時段服務順暢且能持續，有些學者提出建置叢集架構以逹HA（高可用性），惜須較高之設置與時間成本，致此種架構未能普及。為了即時監控許多單機伺服器之運作狀態，本研究提出以叢集架構為基礎，整合雲端概念，建置主機監控系統，並分析不同叢集架構下效能之表現，且由實驗結果尋找出成本、建置門檻與效能兼具之建置架構。
本研究分別以單機、DNS-Base以及Http Redirection之叢集架構，配合負載平衡演算法之輪詢排程法，實際建置雲端主機監控系統，並進一步實務操作，評估分析前述三種架構對主機監控系統效能之影響。研究結果發現以兩台電腦透過輪詢排程法，偵測作業之系統回應時間與系統負載皆優於單一主機，Http Redirection架構系統回應時間比DNS-Base為佳；但主機監控系統之負載則無明顯差異。

關鍵字

雲端計算 ； 叢集架構 ； 輪詢排程法 ； 監控系統 ； 負載平衡

並列摘要

The study constructed a cloud host monitoring system by using Round-Robin Scheduling of load balancing algorithms with cluster frameworks of single computer, DNS-Base, and Http Redirection, respectively. A practical operation was conducted to test and analyze the impacts of these three frameworks to the host monitoring system. The results indicated that the response time and loading of the system worked better when using round-robin scheduling with two computers than with single one. The Http redirection showed better response time than DNS-Based. There was no significant difference in the loading of host monitoring system.
Web services rely on their servers. How to spread the server's load effectively to stabilize the service is an important issue for cloud computing technology. In order to provide service smoothly and consistently during peak hours, some scholars suggested that building cluster frameworks to achieve HA (high availability) could work; however, it is rather costly and time-consuming, so this suggestion is not widely used in general. Due to this fact, this study proposed cluster frameworks as a basis by integrating the concept of cloud computing and constructing a host monitoring system to monitor the operational status of a single server. The performances of different cluster frameworks were analyzed. The purpose of the study was to identify a framework which was able to sustain a reasonable cost, threshold, and effectiveness through the experiments.

並列關鍵字

Cloud Computing ； Cluster ； Round-Robin Scheduling ； Monitor System ； Load Balance

參考文獻

沈炳宏、洪羿漣（2010）。雲端運算智庫–RUN!PC｜精選文章｜開發技術｜初探Hadoop開放原始碼平台環境。上網日期：2010年7月29日。檢自：http://www.runpc.com.tw/content/cloud_content.aspx?id=105318

吳子硯、蘇建郡(2006)。支援SNMP及動態負載平衡之叢集式防火牆設計與建置(碩士論文)。南台科技大學資訊管理研究所。

周天穎、衷嵐焜、辜文元(2010)。格網與雲端技術與GIS。國土資訊系統通訊。

張西亞、王順泰、周朝宜、陳德民、吳長興、李金泓(2007)。高效能電腦叢集的發展與趨勢。物理雙月刊。29(5)，936-947。

陳信宏、包蒼龍(2009)。網頁伺服器負載平衡架構效能之比較(碩士論文)。大同大學資訊工程研究所。