| With the development of the social information, more and more attention has been paidon the protection of the security of the data. HustBackup is a disaster recovery systemwhich was developed to provide data backup for the WAN users. However, multiple users`concurrent backup will cause the server load unbalanced and reduce each server`sprocessing performance. It is an urgent problem about how to effectively balance thedifferences between the various servers, to maximize the use of available resources, sothere is the load balancing technology.The difficulties of load balancing concentrate in the judgment and collecting of the loadinformation. The main load factors of HustBackup are the size of the buckup job and theamount of the backup job. The collection of the load information uses the method ofreal-time dynamic feedback. On the issue of the judgment of the load information, itconsiders several vital factors of HustBackup such as the priority of the user, the size ofthe job, the attribute of the job, and the tactic of the job. In the process of the collection ofthe load information, it includes system starts running load configuration information, thejob load factor after system runs, the load factor of the storage server and the load factor ofthe entire system. The main process of load balancing can divided into two steps. Whenthe load does not exceed the load threshold of the storage server, it uses the way of pollingwith weights to find the lightest load storage server to respond to the backup request andadd the load the storage server. And when the storage server is overloaded, it dynamicallyallocates the server according to the priority of the user and the Revocation of the job.When the users request for restoring job, data can be transported parallel for multiplestorage servers.It has been proved by comprehensive test that, with the use of the load balancing basedon the HustBackup system, the service delay has been greatly reduced, and the load of thestorage service comes to a balanced state, and the resource utilization of the storageservice has been greatly improved. The effect of this algorithm for server load balancing isquite significant, and it improved overall system performance, at the same time ensuresthe quality of service provided. |
