粒子群演算法應用於節能式雲端動態資源配置之研究

A Power Saving-based Dynamic Resource Allocation Using Particle Swarm Optimization for Cloud Computing Environments

陳暉凡(Hui-fan Chen)

指導教授 ： 周立德

國立中央大學/資訊電機學院/資訊工程學系/碩士(2012年)

摘要

雲端運算是近年來熱門的議題，透過虛擬化的機制讓虛擬機器共享同一台實體機器上的硬體資源進而提高實體機器上的資源利用度。然而，在雲端資源配置領域裡，有許多非常重要的議題，如最大效能配置以及目前最令人注目的綠色運算（Green Computing），因為隨著地球資源亮起紅燈，如何以最低耗能成本來完成工作為主要目的。本研究應用於雲端資料中心環境當中，針對低耗能成本問題，考量伺服器以及冷氣空調的耗能，透過雲端資源提供者分別監測伺服器與虛擬機器內部資源的使用率與其剩餘情形，並搭配使用最小平方化的迴歸技術來進行使用率預測，來探討以節能為目標的資源配置策略。本研究提出一個以節能式雲端資源動態配置機制，使用粒子群演算法（Particle Swarm Optimization Algorithm, PSO），來建置一個雲端資源配置策略，簡稱PCRA，滿足使用者的需求以及使得雲端資源耗能較低。PSO是一種基於疊代的優化工具，PSO的優勢在於簡單容易實現、有記憶性、所需的參數設定較少，較適合連續性最佳化問題。本研究利用CloudSim這個模擬器來測試我們系統的演算法效能，實驗的結果顯示，本研究提出之雲端資源動態配置機制PCRA的耗能會隨著伺服器的超載度定義值上升，耗能的改善度也會大幅上升。另外本研究所提出之雲端資源配置機制PCRA對於BFRand之雲端資源配置機制是利用Best Fit 方式來對虛擬資機器與伺服器進行配對，另外搭配random選擇進行虛擬機器的遷移，其耗能改善度為29.98%；對於BFMinU之雲端資源配置機制是利用Best Fit 方式來對虛擬機器與伺服器進行配對，另外搭配選擇最小CPU使用度在需要進行遷移的虛擬機器，其耗能改善度為29.35 %；對於BFMinR之雲端資源配置機制是利用Best Fit 方式來對虛擬資機器與伺服器進行配對，另外搭配選擇最小記憶體在需要進行遷移的虛擬機器，其耗能改善度為30.68 %。因此本論文提出之PCRA能夠提供雲端資源提供者一雲端資源動態調配機制且具有良好的節能效果。

關鍵字

粒子群演算法 ； 最小平方化迴歸 ； 資源配置 ； 雲端運算

並列摘要

Cloud computing is a hot topic in recent years. According to virtualization technology, virtual machines share hardware resources and thus improve resources utilization in the cloud computing. The cloud resource allocation is important issues about how to be an optimal allocation with the lowest energy cost to complete the work. Particle swarm optimization (PSO) is a metaheuristic, it can solve the combinatorial problem likes resource allocation on the dynamic cloud environment. And PSO is more quickly convergence and less parameter setting, so it is easy to implement. To solve cloud resource allocation problem, this paper proposes a Power saving-based Cloud dynamic Resource Allocation (PCRA) mechanism using particle swarm optimization (PSO) algorithm. By using the least squares regression technology, the PCRA can also forecast physical machines utilization. The PCRA allocates virtual machines to obtain the lower power consumption in the cloud environment. This paper use simulation to test and verify the improvement of power consumption. The experimental results show the power consumption of PCRA mechanism compared with Best Fit placement and Random selection (BFRand) mechanism is improved up to 29.98%. The power consumption of PCRA mechanism compared with Best Fit placement and Minimum CPU Utilization based selection (BFMinU) mechanism is improved up to 29.35 %. And the power consumption of PCRA mechanism compared with Best Fit placement and Minimum RAM size based selection (BFMinU) mechanism is improved up to 30.68%. This paper presents PCRA mechanism is able to provide a dynamic cloud resources allocation mechanisms with energy efficiency.

並列關鍵字

resource allocation ； cloud computing ； the least-squares regression ； particle swarm optimization

參考文獻

[1] I Foster, Y Zhao, I Raicu, S Lu, “Cloud Computing and Grid Computing 360-Degree Compared,” Grid Computing Environments Workshop, pp. 1-10, Nov 2008.

[2] R. Yamini, “Power management in cloud computing using green algorithm,” Proceedings of the International Conference on Advances in Engineering, Science and Management (ICAESM), pp.128-133, 30-31 Mar. 2012.

[6] N. Daniel, “The Eucalyptus Open-source Cloud-computing System,” Proceedings of the 9th IEEE International Symposium on Cluster Computing and the Grid, Shanghai, China, 2008.

[11] W. Tian, S. Su and G. Lu, “A Framework for Implementing and Managing Platform as a Service in a Virtual Cloud Computing Lab,” Proceedings of Second International Workshop on Education Technology and Computer Science 2010, Wuhan, China, vol.2, pp.273-276, 6-7 March 2010.

[15] M. Chen, L. Mengkun and C. Fuqin, “A Model of Scheduling Optimizing for Cloud Computing Resource Services Based on Buffer-pool Agent, “Proceedings of IEEE International Conference on Granular Computing 2010(GrC 2010), San Jose, CA, pp107-110, 14-16 Aug. 2010.