計畫名稱：以負載模型推估為基礎的雲端資料中心伺服主機智慧型負載平衡機制的設計與實現--II
執行起迄：2016/08/01~2017/10/31
總核定金額：527,000元
中文摘要：虛擬化技術(virtualization)是雲端運算(cloud computing)服務中非常重要的一個環節，透過雲端運算虛擬化技術，企業將能更有效節省企業資訊維運與發展成本。建立自用或公用的資料中心提供虛擬機的服務與應用，是目前與未來非常重要的雲端運算發展趨勢。負載平衡(load balance)是透過分配工作量到不同的分散式系統伺服器，來確保每台伺服器工作量的平衡，是管理資料中心大量虛擬機重要的機制。若負載調整的相關行動不佳，反而迫使虛擬機管理系統重複進行負載平衡而事倍功半。面對雲端環境應用的多元化，虛擬機的管理機制勢必應朝向智慧型動態負載平衡發展。本研究希望發展雲端運算虛擬機伺服器的智慧型負載平衡機制，透過機器學習方法來獲得虛擬機的負載模型，並以負載模型為基礎發展管理資料中心大量虛擬機的動態負載平衡機制。 本年度計畫從虛擬機的管理參數著手，使用演化式符號迴歸(evolutionary symbolic regression)的技術建立符號式負載平衡模型，並以歸納法建立有效的負載調整配置方式。
英文摘要：Virtualization is an important technique for delivering IaaS (infrastructure-as-a-service) in cloud computing. With the techniques of virtualization, users can reduce costs in building hardware infrastructure and gain flexible and scalable computing resources. Load balancing is to distribute workload across multiple computer servers to achieve optimal resource utilization, maximize throughput, minimize response time, and avoid overload and is also essential in the management of virtual machines. With the variety of applications in cloud computing, balancing workloads of virtual machine servers must be performed dynamically and consider various factors that affect the workloads. Thus, the design of load balancing mechanism for virtual machine servers is a multiple optimization problem. This project focuses on the techniques of computational intelligence for developing intelligent load balancing methods for the management of virtual machine servers in cloud computing. This report presents the analysis of the workload models of single virtual machine server and a prototype using the techniques of evolutionary symbolic regression for load-balance of virtual machines.