The Research And Implementation Of A Hybrid Storage System Based On Data Classification

With the emergence of new non-volatile storage medium, a hybrid storage system have more choices, however, the new non-volatile storage medium compared to traditional hard drives in terms of price, life does not have the advantage. Thus, the current storage architecture will be used at the same time the traditional storage medium and the new storage medium, resulting in a variety of different media composition of hybrid storage systems, in order to take full advantage of the advantages of different storage media and try to avoid its disadvantages. Currently, a hybrid storage system based on different devices has become research institutions and enterprises research difficult and hot.Taking into account the characteristics of different storage devices in terms of price, performance, capacity, based on the full investigation and analysis on the basis of the file system in different types of data I / O request, the research and design of a hybrid storage system can classify different types of data. The paper proposes an I/O classification architecture to close the widening semantic gap between file systems and storage systems. By classifying I/O, a file system can request that different classes of data be handled with different storage system services. At the same time, storage system provides different reliability for different types of data with different levels of RAID. First, the file system layer marks two types of data, namely data and metadata. The storage system can obtain information from a modified file system, thereby distinguishing between data and metadata. Considering the characteristics of random access, small size and high frequency, the paper use SSD to store metadata and HDD to store data to ensure cost-effective system. Using different storage devices to store different types of data, not only can the bandwidth of multiple storage devices be aggregated, but also considering the characteristics of the data to the appropriate storage device, the system performance will be improved as well as cost savings.In stand-alone system and iSCSI network storage, the test results show that a hybrid storage system based on request types enhances the overall performance.