具有负载均衡特性的容错复制技术研究与实现

发布时间：2018-12-11 20:16

【摘要】：随着计算机网络的不断发展及计算机硬件成本的日趋下降，将计算任务分布到不同的物理节点上进行处理，加快任务的处理速度已成为一种研究的趋势。这种分布式计算技术已经在对可靠性要求很高的应用领域取得发展，包括军事、航空、金融、电信等。本文课题来源于“软件应用集成平台容错软件设计与开发”项目，软件应用集成平台面向的应用系统规模庞大，实时性要求高，且应用构件之间相互联系形成复杂的应用系统，其中任何一个构件发生故障都可能导致整个应用系统无法继续正常运行。因此，需要提高应用系统的可靠性和实时性。容错技术通过备份多个副本提高系统的可靠性，组中所有成员都执行相同的操作，系统资源存在浪费，且组成员按顺序执行请求，请求响应时间慢。而负载平衡技术则是将请求分配给不同的成员，由不同的成员同时执行不同的请求，缩短请求的平均响应时间。因此，本文在容错复制技术的基础上引入负载平衡技术，提出了一种适用于请求-状态相关及请求-状态无关的系统模型、具有负载平衡特性的容错复制技术。本文所提出的具有负载平衡特性的容错复制技术在保证组成员状态一致性的前提下，由负载最轻的成员执行请求，实现组成员间的负载平衡。容错组中各成员直接接收客户端发送的请求，且根据本地的负载信息列表决定是否由自身执行请求，其中只有负载最轻的成员执行当前请求。根据请求分配的结果，组中各成员同时执行不同的请求，并直接返回请求响应结果，从而在充分利用容错技术引入的冗余资源的基础上，缩短了请求的平均响应时间，提高系统执行效率。为保证组成员间的状态一致性，执行完与状态相关的请求后，需将改变的状态发送给组中各成员；组成员接收到改变的状态后，更新自身的状态；确认组成员间的状态同步后，再返回请求的响应结果。若请求与状态无关，则直接执行请求，并直接返回请求响应结果。通过实验测试，所提出面向负载平衡的主动复制技术能减少系统资源的占用，，缩短请求的平均响应时间，提高系统运行效率。在执行与状态无关的请求时，组成员直接与客户端交互信息，比存在协调者的容错复制算法减少信息传递的次数，缩短请求响应时间，但由于将协调者执行的操作运行于组中各成员，增加组成员的占用资源；执行与状态相关的请求时，请求平均响应时间随着与状态相关的请求数目的增加而增加。最后，实现基于本文提出的具有负载平衡特性的容错复制算法的容错系统。
[Abstract]:With the development of computer network and the decrease of computer hardware cost, it has become a research trend to distribute computing tasks to different physical nodes to speed up the processing of tasks. This distributed computing technology has been developed in applications with high reliability requirements, including military, aviation, finance, telecommunications and so on. This paper comes from the project of "Software Application Integration platform Fault-Tolerant Software Design and Development". The software application integration platform is faced with a large scale, high real-time requirements, and the application components are related to each other to form a complex application system. Any failure of any of the components may cause the whole application system to continue to run normally. Therefore, it is necessary to improve the reliability and real-time of the application system. Fault-tolerant technology improves the reliability of the system by backing up multiple replicas. All members of the group perform the same operation, the system resources are wasted, and the group members execute requests sequentially, and the response time of requests is slow. Load balancing technology is to assign requests to different members, and different members execute different requests at the same time, so as to shorten the average response time of requests. Therefore, this paper introduces load balancing technology based on fault-tolerant replication technology, and proposes a fault-tolerant replication technology with load-balancing characteristics, which is suitable for request-state correlation and request-state independence. The fault-tolerant replication technique with load balancing property proposed in this paper can balance the load among group members by executing the request by the least loaded member under the premise of ensuring the consistency of group members' states. In the fault-tolerant group, each member directly receives the request sent by the client, and decides whether to execute the request by itself according to the local load information list, in which only the least loaded member executes the current request. According to the result of request allocation, each member of the group executes different requests at the same time, and directly returns the response result of the request. Thus, the average response time of the request is shortened on the basis of making full use of the redundant resources introduced by the fault-tolerant technology. Improve system execution efficiency. In order to ensure the state consistency among the members of the group, after executing the request related to the state, the changed state should be sent to the members of the group, and the member of the group should update its own state after receiving the changed state. After confirming synchronization between group members, the response to the request is returned. If the request is state independent, the request is executed directly and the result of the request response is returned directly. The experimental results show that the proposed active replication technology for load balancing can reduce the resource occupation of the system, shorten the average response time of the request, and improve the efficiency of the system. When executing a state-independent request, group members directly interact with the client, which reduces the number of times of information transfer and shortens the response time of the request, compared with the fault-tolerant replication algorithm with a coordinator. However, the operation performed by the coordinator is run on the members of the group, which increases the resources occupied by the members of the group. When state-related requests are executed, the average response time increases with the increase of the number of state-related requests. Finally, a fault-tolerant system based on the proposed fault-tolerant replication algorithm with load balancing property is implemented.
【学位授予单位】：南京航空航天大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP302.8

【参考文献】