An Experimental Research On The Agent-based Cluster Load Balancing Model

With the rapid development of Internet technology, more and more people begin toretrieve news information and do shopping through the Internet. Many people would like tohave an e-account, even many enterprise make transactions through the Internet in order togain more ratification. But troubles are coming. With the increasing scale of Internet users,the load of the Internet is becoming heavier. Although enhancing the performance of theInternet hardware could somewhat relieve this problem，but the continuously increasingaccess request brings even more load to the servers. So, the development of hardware isimportant indeed, but there is also a need of corresponding load balancing strategy.Fortunately, the Computer Cluster technology preferably solved this problem.Cluster (also called Computer Cluster) accomplish computational task by a group ofloose combined software and hardware who cooperate highly tightly. One of the advantagesof cluster technology is that it could achieve high performance, high fault tolerance and highflexibility with a very low cost. Actually, a large scale cluster could be seen as a whole. Eachserver corresponds to a node, and server could connect to each other in number of ways.Reasonable scheduling among tasks is the core of the cluster technology.Nowadays dynamic balancing load technology is widely used in Internet clustertechnology. The load balancing techniques are often divided into two categories: Dynamicload balance and Static load balance. In order to achieve improved performance of the entirecluster system to enhance the overall network load capacity, improve network response timegoal, this paper, a more flexible intelligent Agent technology, this paper proposes a dynamicadaptive load balancing model based on Agent. And through the different request packetanalysis tasks, divided request task into N categories, each category corresponding to arelatively efficient load balancing schedule algorithm, which have enhanced the performanceof the cluster significantly.This paper mainly accomplished the following works:1.Built an Agent based dynamic adaptive cluster load balancing generalized model byintroducing the Agent technology, which could not only improve the intellectuality of the load balancing system, but also enhance the adaptive capability;2.By introducing the idea of recognition and classification of the request task andcombining the Agent technology, achieved the automatic recognition and classification of thecluster task, which further enhance the operational efficiency of the system;3. Based on the aforementioned model, designed a corresponding algorithm andimplement it. Our model is proven to be effective through the comparison with the originalLVS cluster on several performance parameters;4. Analyzed the special characteristics of the streaming media service and its demands onthe streaming media. On this basis, applied the Agent based dynamic adaptive load balancingmodel to the cluster environment of streaming media servers, and do specific optimizationsaiming at the special characteristics of the streaming media;5. Applied the Agent based dynamic adaptive load balancing model to the clusterenvironment of streaming media servers, and the model is proven to be effective in astreaming media server cluster through the comparison with the original cluster;6. Analyzed in depth the characteristics of the mobile Internet itself, including thedifferences of the user’s behavior, different network throughput, service scale, and servicerequirements. And new challenges brought by these were also analyzed;7. Aiming at the characteristics and new challenges, designed an Agent based dynamicadaptive load balancing cluster for the mobile Internet. The cluster is designed according tothe model we proposed in this paper, at the same time give considerations to thecharacteristics of mobile Internet;8. Constructed the model mentioned above in commercial environment, designed andimplemented4service types to access the cluster. The network throughput, service load andservice demands of these service types are different, which almost cover the common servicetypes of mobile Internet, simple and generalized, could validate our work convectively.In summary, this paper introduce the technology of Agent and the idea of recognitionand classification of the request task,presents an Agent based dynamic adaptive loadbalancing model and analyze the performance of this model; innovation used this model instreaming media, do specific optimizations aiming at the special characteristics and demandsof the streaming media; systematically analyses the characteristics and demands of mobileInternet and do specific optimizations on our model according to these characteristics,implement and validate it in commercial environment. The results showed that the theoreticalsignificance and application value of our Agent based dynamic adaptive load balancing modelcan provide a reference to the same kind of research and application.