Research On The Metadata Management Mechanism In Mass Storage System

The mass storage systems need to be saved Terabyte and Petabyte level even more massive data. The metadata, such as file names, attributes, saved address and access authorization information is generally managed by the metadata servers. Before access to the data of mass storage systems, you need to find and access the metadata. Therefore, the metadata management mechanism is directly related to the mass storage systems' I/O performance.The mass storage systems generally exploit the hierarchy architecture or hashing scheme to manage the metadata, which requires more time and memory consumption for some operations, such as modifying metadata and querying directory, and it doesn't optimize the data access feature ,thus, it seriously affects the I/O performance of the system.This paper introduces DBMS technology and data classification methods based on the analysis of the mass storage system metadata management features to improve the efficiency of metadata management. The concrete works of this dissertation can be summarized as follow:First, we introduce the two-dimension table to preserve the metadata information and propose a new metadata management strategy based on DBMS and then give the various operating processes of metadata. We analyze the time and space consumption and adaptability to different operating environment for this strategy and realize this metadata management strategy prototype system. We collect actual file system metadata and build a variety of test environment for testing and analysis. The results prove that: using two-dimensional table preserve metadata can effectively reduce the time and space consumption, improve the flexibility of metadata management and enhance adaptability.In the analysis of mass storage system access request based on the characteristics of time, we introduce the data life cycle technology and design the metadata classification algorithm and divide metadata into active and non-active metadata. Then, we design the partition index algorithm for active metadata to improve query performance of active metadata and design the improved hash functions index method for non-active metadata to reduce non-active metadata management time and space consumption. Through analysis, we validated that the algorithm can reduce the time and memory consumption for query and index update. We collect actual file system metadata and build a variety of test environment for testing and analysis to realize this metadata index algorithm prototype system. The results show that: the classification index algorithm can effectively improve the performance of metadata query.