【摘要】：近幾年,云計算技術越來越成熟,并被成功運用到了商業(yè)、教育、科研等領域,成為炙手可熱的計算機技術研究熱點。相較之前的Web服務,云計算具有更高的可靠性、擴展性和靈活性,它利用虛擬機、虛擬內存等技術實現(xiàn)了虛擬化,形成了按需支付的模式,為服務提供商和終端用戶提供了諸多方便。在此基礎上,由計算機資源封裝而成的服務數(shù)量不斷增加并被發(fā)布到云計算平臺上為終端用戶提供服務。隨著網(wǎng)絡上可用服務數(shù)量的增加,用戶不再僅僅關注服是否可用,而是更加關注服務的質量(Quality of Service,QoS),比如,執(zhí)行時間、花費多少等。面對任務的繁多和用戶服務高質量的需求,如何對云環(huán)境下的資源和任務進行合理調度成為云計算研究的主要問題之一。另一方面,云計算系統(tǒng)中數(shù)據(jù)中心服務器數(shù)量多、服務資源異構多樣、用戶基數(shù)大、用戶服務約束條件各不相同、應用任務類型各異,云計算數(shù)據(jù)中必須能夠時刻可靠地處理海量用戶任務和數(shù)據(jù),如何及時高效并安全的將結果反饋給終端用戶并能滿足終端用戶的需求成為服務提供商面臨的最大挑戰(zhàn)。同時,完成任務調度所產(chǎn)生的成本也是服務提供商最關心的問題之一。因此,高效的調度算法成為云環(huán)境下研究的重難點。在此基礎上,本文提出了一種改進的蟻群算法,該算法綜合了計算資源(通常指虛擬機Virtual Machines,VMs)的可獲得性以及具有不同服務質量(QoS)約束的任務的特性。鑒于傳統(tǒng)蟻群調度算法一般只考慮計算資源的特性,而忽略了用戶約束條件以及云資源的異構性,本文中的算法將用戶任務分為計算密集型和網(wǎng)絡交互密集型兩種類型,并根據(jù)QoS優(yōu)先級和虛擬機處理速度分別對用戶任務和虛擬機進行排序,旨在能夠在異構環(huán)境中具有不同資源參數(shù)的計算資源上對具有不同服務質量需求的任務進行合理調度,以節(jié)約執(zhí)行時間和成本,同時滿足服務提供商和終端用戶的需求。實驗結果表明,本文中提出的基于預測的調度算法更能傾向于找到最合理的任務虛擬機分配對,并且反復執(zhí)行該算法的情況下能在一定程度上減少任務總執(zhí)行時間和成本。
[Abstract]:In recent years, cloud computing technology has become more and more mature, and has been successfully applied to business, education, scientific research and other fields, and has become a hot research hotspot of computer technology. Compared with the previous Web services, cloud computing has higher reliability, scalability and flexibility, it uses virtual machine, virtual memory and other technologies to achieve virtualization, forming an on-demand payment model, It provides a lot of convenience for service providers and end users. On this basis, the number of services encapsulated by computer resources continues to increase and is released to the cloud computing platform to provide services to end users. With the increase of the number of available services on the network, users are not only concerned about the availability of service, but also more about the quality of service, such as the execution time, the amount of time spent, and so on. In the face of various tasks and high quality user service, how to reasonably schedule resources and tasks in cloud environment has become one of the main problems in cloud computing research. On the other hand, the number of data center servers in cloud computing systems is large, the service resources are heterogeneous, the user base is large, the user service constraints are different, and the types of application tasks are different. Cloud computing data must be able to deal with massive user tasks and data reliably at all times. How to efficiently and safely feedback the results to end users and meet the needs of end users becomes the biggest challenge for service providers. At the same time, the cost of task scheduling is also one of the most concerned issues for service providers. Therefore, efficient scheduling algorithm has become a heavy and difficult problem in cloud environment. On this basis, an improved ant colony algorithm is proposed, which combines the availability of computing resources (usually referred to as virtual machine Virtual machines) and the properties of tasks with different quality of service (QoS) constraints. Since the traditional ant colony scheduling algorithm only considers the characteristics of computing resources, but ignores the user constraints and the heterogeneous nature of cloud resources, the algorithm in this paper divides user tasks into two types: computational intensive and network interaction intensive. According to the QoS priority and the processing speed of the virtual machine, the user tasks and the virtual machines are sorted respectively. The purpose of this paper is to schedule reasonably the tasks with different QoS requirements on the computing resources with different resource parameters in the heterogeneous environment. To save execution time and cost, while meeting the needs of service providers and end users. The experimental results show that the proposed scheduling algorithm based on prediction is more inclined to find the most reasonable task virtual machine allocation pairs and can reduce the total task execution time and cost to a certain extent when the algorithm is executed repeatedly.
【學位授予單位】：華北電力大學(北京)
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：TP3;TP18

【參考文獻】

中國期刊全文數(shù)據(jù)庫 前3條

1 高燕飛;陳俊杰;強彥;;Hadoop平臺下的動態(tài)調度算法[J];計算機科學;2015年09期

2 封良良;夏曉燕;賈振紅;張?zhí)?;實驗基于資源預先分類的云計算任務調度算法[J];計算機仿真;2013年10期

3 房秉毅;張云勇;陳清金;賈興華;;云計算網(wǎng)絡虛擬化技術[J];信息通信技術;2011年01期

中國碩士學位論文全文數(shù)據(jù)庫 前2條

1 郭麗嬌;云環(huán)境下虛擬資源調度方法的研究[D];太原理工大學;2014年

2 范宗勤;云環(huán)境下的資源調度算法研究[D];北京交通大學;2014年

，

本文編號：2153842