Research On Cloud Service Middleware And Its Application In Typhoon Forecast System

On today when Web application is more and more popular, Web server undertakes tasks increasingly onerous, especially for the calculation of the large amount of data, the resource of the server’s hardware becomes bottleneck of the efficiency of computation. However simply improving the performance of the server itself doesn’t make much effect and there exists a waste of resources. If the computing tasks can be broken down into several sub-tasks, assigned to the idle computing servers, then because of the advantage of parallel computing, the efficiency will be greatly improved.Cloud computing technology can solve the problem of efficiency of large tasks. Cloud computing is a kind of commercial computing models, it can break down computing tasks into smaller tasks, distribute them to a large resource pool constructed by computers so that a variety of applications can obtain necessary computing ability, storage space and information services. As the computing process of cloud computing tasks is carried out locally on each computer where only task description and computing results are transported on the network, the cost of the network transmission is very little. The efficiency of computation depends on the average performance of compute nodes and load balancing algorithm.This paper proposes a kind of cloud service middleware framework based on Multi-Agent, divided into three levels, the user interface layer, SOA layer and resource management layer. User interface layer is responsible for the description of the task interface, Web request node calls computation under cloud service mode through the interface. The SOA layer is service-oriented. It manages services provided to the user interface by cloud service middleware, maintains the list of services and can find the appropriate service according to the user’s request. Resource management layer is the core part of the entire cloud service middleware. It contains a load balancer and a monitor providing fault tolerance service and log service, achieves the logic of decomposition and distribution of tasks and summary of results and maintains a list of compute nodes.A complete Agent Platform consists of several Agent containers, under each container there are several agents and a massage scheduling module which Agents use to send and receive message for communication. In the Cloud Service Middleware of Typhoon Forecast which is called CSMTF for shorted there is an Agent responsible for task decomposition, distribution and summary of the results called master node. The master node receives computing tasks from the request node with a load balancer inside real-time getting each idle computer’s performance and assign tasks to lightly loaded nodes. The log monitor in master node is responsible for the collection and recording of the Agent log so that managers view to understand the work of Agents. At the same time, master node has fault-tolerant monitor inside to detect Agent running and restore the Agent which goes error. In the Cloud Service Middleware there are several agents responsible for computing of sub-tasks called computing node. Computing nodes send real-time native load information to the master node, receive sub-tasks distributed from the master node, compute locally and return the result to the master node. Computing nodes also have a log monitor used to record the information of the local Agent, the fault-tolerant monitor inside used for local recovery.The typhoon forecast system is a typical Web application. It is a consulted Group Decision Support System which is called GDSS for shorted. In a case experts forecasting routes involving large amount of data computation. Web servers need to calculate similar path for forecast routes and all the historical typhoon routes and it requires real-time display of the most similar route. However, there are thousands of points of the route in the database. It is difficult to compute efficiently in the case of the Web server with limited resources. This paper applies the technology of Cloud Service Middleware to the typhoon forecast system in order to improve the efficiency of computing similar route.At last, this paper builds a lab environment for testing performance of the CSMTF system, compared to Web server’s single computation the computing speed has been improved significantly and the response time of the server has been greatly reduced.