基于ILink平台的负载均衡调度研究

发布时间：2018-01-27 08:16

本文关键词： ILink平台负载均衡过载感知动态负载均衡过载迁移　出处：《中国民航大学》2017年硕士论文　论文类型：学位论文

【摘要】：近年来航空货运量逐年增加,加大了航空货运数据交换平台ILink的负载压力。目前ILink平台部署的是静态加权轮询算法,但该方法随机性大、容错性能差,后台节点负载信息收集不及时不全面,且过载阈值的设定是静态不变的,不能根据服务器各个时刻的负载状态进行动态调整,容易造成部分节点负载过重甚至宕机,需要进一步对该负载均衡方法进行改进。首先,提出了一种链网结构的软件负载均衡模型,其思想是组内采用网状模型,组间采用链状模型,通过命令驱动和负载状态驱动相结合的方式快速收集节点信息。该模型既能保证集群中服务器节点的信息共享又能保持系统的高可靠性,为动态负载均衡算法的提出奠定了基础,有效解决了负载信息收集问题。其次,针对航空物流ILink数据平台交易复杂性高、服务种类多、业务规模大等特点,结合四种基本服务类型的特征,提出一种面向多类型服务的动态负载均衡算法。该算法通过实时采集四种服务类型的资源消耗,并根据用户服务请求类型,动态调整各服务器的负载,同时引入过载服务器节点缓冲机制缓冲负载过重节点,有效解决了航空货运数据交换平台ILink可能出现的用户等待时间过长、局部负载过重、甚至个别服务器节点崩溃的问题,达到集群的负载均衡。最后,提出了一种负载感知的自适应阈值过载迁移策略。该策略根据用户请求分发到服务器节点的速率,自适应调整过载阈值;采用负载感知策略,迭代选择待迁移服务器中负载率较高、迁移代价较小的服务为迁移集合。该方法降低了系统的负载率,提高了约1.2%的吞吐量,进一步提高了系统的可靠性、资源利用率,改善了通信性能。
[Abstract]:In recent years, the volume of air cargo increased year by year, which increased the load pressure of air cargo data exchange platform (ILink). At present, static weighted polling algorithm is deployed in ILink platform, but this method is stochastic. Fault tolerance performance is poor, the background node load information collection is not timely and comprehensive, and the overload threshold is static unchanged, can not be dynamically adjusted according to the load state at each time of the server. It is easy to cause some nodes to be overloaded or even down, so it is necessary to further improve the load balancing method. Firstly, a software load balancing model with chain network structure is proposed, the idea of which is to adopt the netlike model in the group. The chain model is used to collect node information quickly by combining command drive and load state drive. This model can not only guarantee the information sharing of server nodes in cluster but also maintain the high reliability of the system. It lays the foundation for the dynamic load balancing algorithm and effectively solves the load information collection problem. Secondly, aiming at the high transaction complexity and the variety of services of aviation logistics ILink data platform. Based on the characteristics of four basic service types, a dynamic load balancing algorithm for multi-type services is proposed, which collects the resource consumption of four types of services in real time. According to the type of user service request, the load of each server is adjusted dynamically, and the overload server node buffer mechanism is introduced to buffer the overload node. Effectively solves the air cargo data exchange platform ILink may appear the user waiting time too long, partial load overweight, even individual server node crashes, achieves the cluster load balance. A load-aware adaptive threshold overload migration strategy is proposed, which adaptively adjusts the overload threshold according to the rate of distribution to the server node. The load aware strategy is used to iteratively select the migrating set as the service with high load rate and low migration cost in the server to be migrated. This method reduces the load rate of the system and increases the throughput of about 1.2%. Further improve the reliability of the system, resource utilization, and improve the communication performance.
【学位授予单位】：中国民航大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：V353;TP301.6

【参考文献】