The Solution Of Disk Load Balancing Based On Disk Striping With Genetic Algorithm

Today is an information age, and information is becoming important to people more and more. With the rapidly increasing of information data, the topic of how to store safely, transfer lately, recover rapidly and how to avoid large loss because of losing data is cared for by more and more people. So the technology of storage has become very important. Because there are many applications based on networks with the high development of Internet, the storage, as the kernel of background devices, must supply the applications with all sorts of services involving the large storage capacity, the cluster of high applicable database, the backup and recovery, the network storage, the high performance compute and so on. In other words, the storage must supply the dependable manipulation and protection for the data in 24×7. The disk mirror, the store buffer and the RAID technology are the most technologies at present.With the increase of the number of disks in storage subsystems due to rapidly increasing capacity requirements, the disk array is formed. In a large disk array, people use the RAID technology for increasing reliability and I/O performance. In the array, a disk group is composed of some physical disks, and the disks that display in front of the client system are called logical disks. Every logical disk is a mapping of a part of the disk group. Because there are some differences of the I/O characters of the application systems on different logical disks, the load of each physical disk will be imbalance possibly. Based on the above-mentioned, the largest performance problem of the storage is load imbalance.In this paper, a new scheme based on Disk Striping and Genetic Algorithm to solve the problem is presented, including the file partition algorithm based on Disk Striping and the file allocation algorithm for load balance. In Disk Striping technology, the realization of the disk load balancing is depended on application programs or file systems. Based on the idea of parting files, a load balance from the point of view of store device is given. On the other hand, Genetic algorithm is a new algorithm to solve an optimization problem. In the Genetic algorithm, it executes three same and simple genetic operators: selection, crossover and mutation. In this Algorithm, an efficient searching is executed in complex spaces leaded by fitness value, until acquiring the best result. There are three genetic operations: selection, crossover and mutation. The majority of searching is completed by selection and crossover, and the ability of approaching the best result is improved by mutation.In this Algorithm, the chromosome of the tree is represented with the tree structure coding. In the design of crossover, we first merge two parent trees to a graph, and then create a new tree from the graph. Based on the method, the algorithm has a good performance of crossover. Our simulation results show that the proposed algorithm is able to find the result with fast speed.Computer simulations were conducted to evaluate the performance of the algorithm in the end of this paper. The results show that the proposed algorithm is a correct and effective algorithm. So it's considered as a good prospect in storage.