| As the increase of the user number as well as the accumulation of the user data, the data stored and administrated in the storage system will grow certainly, data organization and administration appear very important. The proper data organization can improve storage system's performance, and make the system be reliable, available and scalable.By studying the data distribution in the common storage systems and its feature, the research on the adjustment of data organization in the process of system scalability is done. The defection in its scalability is analyzed by studying the characteristics of the storage space and data distribution, and presents a data reconstruction mechanism.Storage system's scalability, which leads to data reorganization and reduces the availability of the system, is a main reason why the cost of system administration increases. By analyzing the data distribution algorithm in the storage system, considering the advantage of data insinuation table and insinuation function, data reconstruction mechanism is presented in the sight of data distribution and metadata's consistency.The reconstruction mechanism takes two factors under the consideration. The first factor is the data distribution characteristics, and the second is the consistency of metadata in the system scalability. Abiding by the data distribution characteristics of the cluster storage system, the mechanism transfers the data from the specific devices to others, and adds new variables to identify the scalability times of the file by using an index table recording the environmental features (before and after the scalability). It shows that the mechanism can eliminate the mess introduced by the file accessing before and after the reconstruction, and guarantee the logical consistency of metadata.It shows that the impact taken to the system performance by data reconstruction mechanism can be controlled in the minimum limit. Besides, the new assume and the future works needed to do are also brought on. |
PDF Full Text Request