The Research Of The Improvement Of Performance For Object-Based Storage Device

With the development of Internet at a full speed and the appearance of requirement of petabyte scale storage system. Object-Based Storage System,which adopts full expressive object interface for storage interface instead of block interface and file interface and avoids the disadvantage of those two interface,becomes the mainstream architecture of massive information storage system.Object-Based Storage system is composed of Client,Metadata Server,Object-based Storage Device and network.Object-based Storage Device is the kernel component of Object-Based Storage System,it is not only the carrier of data,but the organizer and manager of data.It organizes the data in the form of object, and serves the Metadata Server and Client with object interface.Data flows between Client and Object-based Storage Device, performance of the whole Object-Based Storage System is affected directly by Object-based Storage Device.So improving the performance of Object-Based Storage System by taking full advantage of the characterist of Object-based Storage Device will be more and more important.Much researchment has been put onto the OSD-2 working draft putting forward by T10 Technical Committee.By mapping flat namespace to hierarchial namespace which can be a better way to organize objects in Ext2,we exploit to the fully ext2's favourable conditions and avoid unfavurable ones.According to the scheme for disk allocation of Ext2,the distribution of data of file is not discrete but partial continuous. The reconstruction of the index of data block can not only downhold the space occupied by the logical number of data block,but decrease the frequency of disk I/O that results from obtaining the logical number of data block before OSD reading big object. thereby improve the performance of the reading big object of OSD. According to the fact the quantity of data I/O is very large, but the frequency of attribute I/O is very high.Pingpang scheme was employed by data buffer, not only decrease the frequency of disk I/O, but lessen the waiting time of disk.we can make sure the attribute of data in data buffer exist in the attribute buffer by building up relationship between data buffer and attribute buffer. thereby improve the performance of access.