Load Balancing Algorithm Based On Mobile Agent

At present, with the development of information technology, especially network andmultimedia technology, distance education system has been widely used. As the extent of theapplication of remote experiment teaching platform continues to improve and increase the amountof access that appears will impact on the server response time and network service quality. Singleserver processing capacity is limited, it may become a new bottleneck in network access. How todesign a server cluster load balancing and scheduling policies prediction algorithm to solvevarious problems of distance education platform, as well as service platform better, faster inresponse to user requests is the main content of this paperCurrently, centralized load balancing control is widely used. The disadvantage of theincreasing of the cost of network communication as well as the low self-intelligence of theserver individual and great dependency of the node is should not be neglected. Mobile Agents inP2P network is widely, showing good intelligence. Mobile Agent technology in remoteexperimental teaching system combines predictive load balancing algorithm, can be a goodsolution to collect, predict and adjust the load balancing strategy in information problems.Our group has completed the basic functions of distance learning platform design, which isbased on B/S (browser/server) access nodes. Users can remotely log off-site to complete therelevant studying by network. Mobile Agent is applied in the new structure of B/A/S mode in thearticle. The agency layer can be a system medium to complete a load balancing. The use of mobileagent migration algorithm for load collecting and updating the cluster nodes and load informationsystems to achieve LPMA mobile agent subsystem main functions are the main task of thedevelopment and implementation in the article.The mobile agent B/A/S three-tier architecture model is: Underlying-OpenMosixdistributed system platform, to obtain data; Intermediate layer-Aglets mobile agent platform,which includes tasks Agent, integration Agent, task management and scheduling Agent, executeAgent, real-time data collection and sharing knowledge of several parts, such as Agent, Functional intermediate layer is to load the data collection and updates, as well as the migration taskscheduling nodes; Upper-user interface, is mainly used for interface in response to clientrequests.Load collect and update information from real-time data collection is mainly realized byinteracts with the Executive Agent achieve, using polling mechanism to periodically obtain thedata and calculate the size of each load node to determine the need for load migration. CPUutilization, memory utilization, network utilization, I/O utilization, and the current connectionspeed of the server is the acquisition targets. About node migration algorithm, we use theclustering analys is method and random intervals, overload node to select the appropriate nodematches and migration and interactive tasks, so as to achieve load balancing purposes. In theimplementation process, the mobile agent needs to achieve the following three functions:(1)dynamic routing,(2) multi-agent parallel operation,(3) an asynchronous task execution. Thisarticle also gives details on the IBM Aglet platform, using Java language development process.In order to verify the reliability of the superiority of the system architecture and algorithmmodel, a simple heterogeneous distributed systems using virtual machines is established in theLAN environment, and OpenMosix is installed on the system platform to do a series of simulationexperiments. According to the degree of load balancing calculation of the LBMA system to makea comprehensive evaluation. By results of the statistical analys is showed through node selectionand random intervals, can effectively redistribute load migration tasks, making the balance of thesystem significantly improved resource utilization and thus the whole system is correspondinglyincreased. Therefore, if the system is used in distance learning platform can effectively solve thehigh traffic load distribution under the problem of uneven, thereby ensuring real-time and distancelearning intelligence platform.At the end of the article, on the development of this platform are summarized system, andfuture research directions outlook.