| Data is divided into two kinds, one kind can use two-dimensional table stores, using a database management system for data records of the storage management level; another kind is based on file which is directly managed by the file system Usually in the process of data processing, because handling of file is more than the operation of the records and the files which need to store are also more than the records significantly, it puts forward the higher requirements about the I/O performance for storage systems and storage capacity. At the same time, as the data is not born equal, with the passage of life cycle, the actual application value for the data has been reduced and the stored value is also changing. Due to the I/O throughput for storage devices, capacity and the price are different, if not reasonable allocation of storage devices can cause waste of storage resources, increase the storage costs. How to solve the large file level data storage, improve the utilization rate of all storage resources in a storage system and solve the bottleneck problem about the traditional storage in the I/O and storage capacity, hierarchical storage has an important significance. In the existing hierarchical storage system, mostly are for fixed application scenarios, but the importance of data is dynamic, the actual storage value is also different in different applications. Hierarchical storage means that data will be live migration between all levels of the storage device, reasonable data migration needs to follow certain migration strategy, in order to realize the value of the hierarchical storage, for existing migration strategy, mostly there are considering data attribute is not comprehensive. Aiming at the problems existing in the hierarchical storage, based on analyzing research situation and existing problems for the hierarchical storage, this article designs a class file data storage model. Main content is as follows:1. Design a file data access characteristics form of data structure, is used to record data access information. The statistical method for additional access frequency statistics is presented. Getting file access information, the system does not need to scan the entire storage system, only incremental scan record, for gaining the access information.2. Transfer object selection strategy is proposed, mainly according to the properties file access record gain access situation, calculation data of the actual use value, to determine what data needs to be migrated.3. The visit frequency is presented as the main index, heat storage space left and I/O as auxiliary indexes of multi-index data migration strategy to implement the data migration, and how in many policy adaptive to choose the appropriate strategy.4, in the process of data migration, need to reasonably control the data migration process. In external interrupt request, should adopt the corresponding interrupt handling process, to ensure the migration process to continue. Experimental results on the one hand shows the hierarchical storage can effectively solve the storage capacity of the bottleneck problem, and makes the storage device the remaining space in the controllable scope, ensure that takes up space distribution balance, improved utilization of storage resources fully. On the other hand shows applying data migration control applications access to interrupts can improve the system throughput. |
PDF Full Text Request