Based On Apache Cluster Load Balancing Research And Implementation

With the continuous development of Internet and wide applications, makes it the quality of network services have become increasingly demanding, the network server experiencing a big challenge, a lot of network services because the network load of the linear growth overwhelmed the service response time increased, quality of service compromised, cluster load-balancing technology, the solution to server-side bottlenecks in a forceful measures, with high cost performance and scalability, it has won a wide range of applications.Cluster load balancing load balancing algorithm is the key to the node devices by taking into account various factors to make a rational and balanced distribution of tasks to the appropriate node in the real device, reducing the response time of the cluster system to increase the throughput of the cluster system to enhance the flexibility and availability of network equipment to increase network communication bandwidth. Cluster load balancing algorithm for the design of the pros and cons will directly affect the performance of the cluster. Dynamic feedback is a server-based real-time load and respond to the situation, constantly adjusting the proportion of requests between servers to avoid some of the server overload is still receiving a large number of requests, in order to achieve increased processing power of the entire cluster.In this paper, the system of cluster and current mainstream introduction of load balancing technology theory and research, for its shortcomings and put forward three kind of new viable solutions, according to dynamic feedback algorithm and the proposed new algorithm, not only take into account the node device performance and load capacity, but also based on the input indicator to adjust the weights of each node, so that load balancing does not tilt. This paper also uses neural network BP algorithm, queuing theory algorithms to improve the load balancing algorithm to make full use of neural network BP algorithm and queuing theory model for a variety of advantages, without human intervention to build a self-learning of the load balancing algorithm, this paper carried out described in detail. Finally, through comparative analysis of simulation experiments show the superiority of the improved algorithm, effectively improve the performance of cluster system, reduce system latency and improve system throughput and utilization.