The Research And Implementation Of Dynamic Load Balancing Strategy Based On Lvs Cluster

With the development of network services, the user visits and the amount of load on the server presents exponential growth, and how to provide users with high-quality, uninterrupted network services, become the problem that network service providers must solve. The use of one single server architecture is difficult to meet this demand, while the cluster technology makes the problem solved easily.Cluster is composed by a set of computers which is independent and interconnected through the high-speed network, and the cluster is managed by a single system mode, which makes the cluster act as a stand-alone server, when interacting with the user. Linux virtual server (LVS) as one of the cluster implementation techniques has been widespread concerned and applied.Although the LVS cluster system is widely used, the load balancing of LVS cluster is not perfect. And the several existing load balancing algorithms can not adjust the load allocation strategy timely according to the real-time load of servers, which is easily to cause the load skew of the system and reduce the cluster performance seriously.To solve this problem, this paper proposes a dynamic load balancing strategy which could adjust servers’weights dynamically, avoid the cluster’s load skew phenomenon effectively, and thus enhance the performance of the whole cluster.The paper mainly involves the following work:First, study the cluster system and load balancing technology to have a preliminary understanding of them. Then, research the LVS cluster technology deeply, and have a more profound understanding of its architecture, working principle and three operating modes. Besides study the existing eight load scheduling algorithm, focus on the analysis of the WLC algorithm, and summarize its existing inadequacies.Next, according to the defects of the WLC algorithm, propose the idea of dynamic load balancing strategy, design and implement the corresponding dynamic load balancing module (Dynamic-Load Module).Then, set up the simple test environment, use the stress testing tool---Web Application Stress Tool(WAS) to execute the performance testing of the original WLC algorithm and improved dynamic load balancing strategy, and analyse the experimental results, which preliminary verifies that the dynamic load balancing strategy can effectively enhance the performance of the cluster system.Finally, use the deployment environment of the China Mobile Research Institute PDM-PAAS project to execute the performance testing of dynamic load balancing strategy under the complex environment. Also use the stress testing tool---WAS to execute the performance testing of the original WLC algorithm and improved dynamic load balancing strategy, and analyse the experimental results, which further verifies the effectiveness of the dynamic load balancing strategy. And the dynamic load balancing strategy can be used in the PDM-PAAS project.