Research on Load Balance based on Fujian Meteorological Cloud Platform

Abstract

Load balancing problem is one of the hot issues in computing. Based on Fujian meteorological platform, the establishment of the model carries out quantification the load status and cluster resource utilization rate of virtual machine, and the improved LC algorithm is proposed with the meteorological data features. The verification shows that this algorithm can improve the load balance of resource utilization and computing. Introduction In order to better meet meteorological, water conservancy, aviation, forestry and other needs, this paper can realize the meteorological disaster monitoring of the rural areas, coastal areas, major rivers, major strategic economic zones and geological disasters prone areas, to improve the comprehensive monitoring and early warning, resilience and disaster reduction of the weather disaster [1,2]. At the same times, it can avoid aviation, water conservancy and other departments on weather radar observation network redundant construction, to maximize the investment benefits. In the province, we can share weather radar observation network information, Fujian Provincial meteorological bureau builds a new generation of data center platform. Load balance is the basic element of computing, and it is the key link of server cluster. Compared to traditional single process load balancing framework, a multi process architecture can take full advantage of parallel processing ability to improve overall system performance in [3-5]. Load balancing technology is a kind of strategy, it can make multiple servers or multiple links shared some heavy computation and I / Q task, so as to lower the cost of eliminating the network bottleneck and improve the reliability and flexibility of the network. Load balancing technology not only can maintain the load balance distribution in the network system, but also can maintain the efficient operation of the network system, so it is the important technology to guarantee the high performance of the network system. In this paper, we mainly study the parallel processing efficiency of the cloud computing with the appropriate load balance algorithm in the special background of meteorological data. Load Balancing Mechanism and its Algorithm under the Cloud Computing Load balancing problem description under the computing. Cloud and core concept are resource pool, and scheduling algorithm is the core of resource allocation, scheduling is load balance between the virtual machine in data center. Cloud task allocation is divided into three levels: task request layer, resource management, task execution layer. The task request layer is facing the user, realizing the user and system interaction obtain the user's request; resource management layer cuts into several sub tasks with logic independent by using the MapReduce tasks, and then sub tasks assign to appropriate the physical machine virtual node according to the load balancing mechanism to parallel processing [6]. This paper only discusses the situation of independent parallel processing after cutting sub tasks; resource management layer is the key to achieve reasonable scheduling of resources. Under the environment, the physical machine is cut into a plurality of virtual 4th International Conference on Computer, Mechatronics, Control and Electronic Engineering (ICCMCEE 2015) © 2015. The authors Published by Atlantis Press 40 resources by using virtualization technology, and the virtual resource is on-demand distribution and automatic growth, but the distribution amount and automatic gain reduction cannot exceed the physical machine itself limit [7]. In a cloud, the number of virtual nodes is segmented, parallel processing resource management layer divides the same number of sub tasks, and it is asked to perform the task execution time as small as possible. Load balancing mechanism has a fundamentally different in the platform and the cluster environment. In a cluster environment, the task processing is based on the actual physical host, if the acquired host is too much, it is not only invested large; when the load is too low, it will cause resource waste serious; if host is too little, load exceeds the total load capacity, it will cause great impact on QoS system, so the traditional load balancing is a rigid load balancing. The platform virtual resources are a dynamic resource environment, which can add or reduce the virtual machine according to the actual load change. Selection load balancing algorithm. There are many load balancing algorithms that are round robin algorithm (RR) algorithm, weighted round robin (WRR) algorithm, least connection (LC) algorithm and weighted least connection (WLC) algorithm [8]. (1) RR algorithm: the customer request can distribute to each member server, RR mode can be simply interpreted as a random choice. This algorithm takes the switch server as the same and does not consider each server connection number and response time, and it is equivalent to randomly select the server of business in the serve. However, this algorithm is randomly chosen for server, so it is not ideal for server load balancing in some cases. (2) WRR algorithm: based on the RR algorithm, according to establish of the operation speed of each server and the number of connection different abilities in the server group, the weights of each server in the server group are added to the response, and then according to the weight of the processing to improve the algorithm load balancing ability. WRR algorithm is suitable for each server with different processing capability. In the algorithm, each server is assigned a weight, the weight is an integer, it is shown that the corresponding to the processing power of the server. In the server set up or release the connection, the server weight values will change. The WRR algorithm chooses the best response server to provide service according to the different weight of each server, which can simulate better server load balancing process. (3) LC algorithm: network connection point have at least the number of connected servers, this is a dynamic switching algorithm, which needs to dynamically calculate the number of connections established by each server, and then to dynamically select the corresponding server according to the obtained results. In the virtual server, all server processing ability and the establishment of connection difference are smaller, if the LC algorithm in the number of requests varied greatly, it can very good request distribution, this is due to the arrival of switch requests will not be pointed at a single server caused by overweight load of the server. However, the LC algorithm does not consider the load imbalance caused by the server performance difference with different processing capabilities servers, (4) WLC algorithm: it is the improvement of LC algorithm. In the WLC algorithm, each server will have a weight value, the server weight value is higher, and indicating that the number of established server connections has lower percentage of the total number of established server connections, server has not many connected and still has considerable service ability. The switch gives the weight for each server, the switch is according to the weight value, the network connection is divided the server. WLC algorithm is to select the server according to established server connection number and weight value ratio, the algorithm finds the smallest ratio server to establish a connection, this ratio shows the current server load condition. Construction of the Load Balance Model under the Fujian Meteorological Cloud Platform Fujian meteorological platform. Fujian meteorological platform is an IaaS platform of heterogeneous resource type, it not only can support access types Web services, but also support data query service based on information service as the core [9]. For the consideration of the application, the platform resource selection decides that the platform needs the server of the X 86 schemas and the server of the RISC architecture, so it is an IaaS platform of heterogeneous