基于事务调度与数据放置的复杂云服务系统负载均衡问题研究

发布时间：2018-01-14 05:25

本文关键词：基于事务调度与数据放置的复杂云服务系统负载均衡问题研究　出处：《山东大学》2017年硕士论文　论文类型：学位论文

【摘要】：由于云计算广为接受,越来越多的传统应用转移到云环境中。云服务系统是部署在云中,按照云计算软件即服务的思想对外提供业务服务的大规模在线应用,如阿里巴巴交易网站淘宝网。云服务的特点为按需服务和随需扩展。大型云服务系统的特征一是用户规模大,请求变化多样,请求热点偏移大;其次是数据规模大,数据存储多中心分布。因此负载不均衡的情况在云服务系统中普遍存在。为了满足用户的需求,云服务系统预留了大量的空闲资源,这使得系统的利用率很低。通过深入分析,云服务系统中的负载均衡显得尤其重要。云服务系统中负载均衡技术成为了当前和未来计算机领域的重点研究问题之一。本文通过深入分析,云服务系统产生不均衡的两个重要影响因素有请求复杂和数据存储不均衡。由于事务的复杂多变情况和数据的放置情况使得云服务系统出现负载不均衡的情况,这极大的影响了用户的体验。因此本文提出了事务调度和数据放置两个方法来进行复杂云服务的负载均衡。本文的主要研究内容有以下两点:1.复杂云服务系统负载均衡中事务的调度方法。在事务调度方法中,复杂云服务系统中由于复杂事务的存在,云服务系统的性能,比如响应时间,会产生很大的波动,这极大的影响了用户的体验。本文通过实验的方法详细分析了引起云服务系统性能波动的原因,并且给出了一个复杂事务的调度方法,用来平缓云服务系统性能的波动情况。2.复杂云服务系统负载均衡中数据的放置方法。在数据放置方法中,复杂云服务系统中数据不是放在一个服务器节点上,分布式的数据存储正是云服务系统的一个特征。数据放置的策略将会直接影响到云服务系统的性能,本文关注了云服务系统中多维相关数据放置问题,提出了一个数据放置的方法。以前的工作都集中于云数据库中多维数据的索引,旨在提供快速查找服务。而本文的工作目标与之不同,本文旨在针对数据的相关性来促进检索处理。本文给出了一个数据分组的成本计算方式,并提出了一个新的得到近似最优数据分组策略的方法。本文搭建了云服务系统负载均衡实验环境,用来验证本文提出的云服务系统负载均衡方法。本文在实验中首先验证了复杂事务引起响应时间波动现象的存在并通过了大量的实验验证了本文提出的平缓响应时间波动的复杂事务的调度算法有效性。另外本文提出了一个数据放置的分组方法,并同样通过实验证明本文的数据放置算法是行之有效的。通过以上两个方面证明了本文提出的云服务系统负载均衡方法框架的准确性。
[Abstract]:As cloud computing is widely accepted, more and more traditional applications are transferred to the cloud environment. Cloud service system is a large-scale online application deployed in the cloud and providing business services according to the idea of cloud computing software as a service. For example, Taobao, Alibaba transaction website. Cloud service is characterized by on-demand service and on-demand expansion. Large cloud service system is characterized by large scale of users, variety of requests, and large deviation of hot spots. Secondly, the data scale is large and the data storage is multi-center distribution. Therefore, the load imbalance exists widely in the cloud service system. In order to meet the needs of users, the cloud service system reserved a large number of idle resources. This makes the utilization of the system very low. Load balancing in cloud service system is especially important. Load balancing technology in cloud service system has become one of the most important research problems in the field of computer at present and in the future. The two important factors that affect the unbalance of cloud service system are the complexity of request and the imbalance of data storage. Because of the complexity of transaction and the situation of data placement, the cloud service system has the situation of load imbalance. This has greatly affected the user's experience. Therefore, this paper proposes two methods of transaction scheduling and data placement to balance the load of complex cloud services. The main contents of this paper are as follows:. 1. Scheduling of transactions in load balancing of complex cloud service systems. Because of the existence of complex transactions in complex cloud service systems, the performance of cloud service systems, such as response time, will fluctuate greatly. This has greatly affected the user's experience. This paper analyzes the causes of cloud service system performance fluctuations in detail through experimental methods, and gives a scheduling method for complex transactions. The fluctuation of performance of a cloud service system. 2. Data placement in load balancing of complex cloud service systems. In data placement methods. Data in complex cloud service system is not placed on a server node. Distributed data storage is a characteristic of cloud service system. The strategy of data placement will directly affect the performance of cloud service system. This paper focuses on multidimensional data placement in cloud service systems and proposes a method of data placement. Previous work focused on the index of multidimensional data in cloud database. The purpose of this paper is to provide a fast search service. The purpose of this paper is to improve the retrieval processing for data correlation. This paper presents a cost calculation method for data grouping. A new approach to approximate optimal data grouping strategy is proposed. A load balancing experimental environment for cloud service systems is built in this paper. In this paper, we first verify the existence of response time fluctuation caused by complex transactions, and verify the proposed flat response through a large number of experiments. The scheduling algorithm of complex transactions with time fluctuation is effective. In addition, this paper proposes a grouping method for data placement. It is also proved by experiments that the data placement algorithm is effective, and the accuracy of the proposed load balancing framework for cloud service systems is proved by the above two aspects.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.09

【相似文献】