| With the development of Internet, some servers are overload because they have to cope with tens of thousands of access in a very short time. How to achieve a reasonable volume of distribution in these network equipment that deal with same work,so that it will not occur the situation that one equipment is busy and other equipments are not,so server cluster load-balancing mechanism is coming.The load-balance scheduling algorithm is the core of the server cluster load balance technology.Usually we select the periodic method to obtain server's load in adaptive load-balance scheduling algorithm, this means that the load value does not represent member's current load, but the past.If there are many customers to link in a short time, the actual load of servers will increase suddenly before the load value is renewed.In view of the above questions,this thesis proposed a adaptive load-balance optimization algorithem based on genetic algorithm,which not only considered the past and current performance of system to server's choice and the assignment, but also estimated future load of system before pre-assign the concurrent request.This thesis mainly studies the fllowing: Firstly,making referance to the model of the network level and classifing the current server load-balance technology.Each working principle of load-balance method was introduced in details,then comparing and analyzing the strengthens,weaknesses and apation of these ways.Secondly,I did an in-depth research on the common load-balance algorithm and drew flow chart of each algorithm,pointing out their characteristics and limitations. Setting about from improving the mathematical model of load balancing,proposing a adaptive load-balance optimization algorithm based on genetic algorithm,and analysising and simulating the algorithm.Finally,we designed a load balancing system based on the platform of windows,which verified the optimization algorithm, then analysed the test results and compared them to other's. |
PDF Full Text Request