General RAID Disk Array Performance Analysis And Optimization

With the rapid growth of the social economy and the rapid development of computing technology, low-speed disk access technology is difficult to meet the user’s demands, which makes the rapid development of cloud storage technology.To enhance the efficiency of access, most of servers based on disk array use RAID technology.RAID technology enhance the I/O operation efficiency and improves the performance of storage systems.On the basis of RAID technology, how to optimize the system to make it as far as possible improve data processing ability,and the work is becoming a research focus.RAID is a virtualization technology, and data remains on disks.There are repeated frequently between disk and the system I/O operations.Mechanical movement greatly slow down the machinery I/O performance and increase CPU overhead.To enhance the I/O processing efficiency, the use of RAID cache system to optimize read and write performance.In the process of reading the data, there is a data block read latency problems.By analyzing the internal structure of the data, the data structure is single, which can not make full use of. So it needs to increase data cache space to optimizate. The core of the strategy is to first save the data in the cache block header BH, when dealing with I/O, the data exchange data directly through the in-memory cache of data,thereby enhancing the ability to access the data read operation. Through test analysis indicates a request to read cache system has significantly improved in many aspects of data striping, RAID block or processes, etc.During the write performance optimization, there is a small data performance issues. By studying the kernel code, we found that small blocks of data will not stop seeking while writing, and this is called I/O wait phenomenon. It uses delay write to optimize performance in the cache. The core of this strategy is to let the small data stored in the cache, and then the small data will aggregate into large data blocks,then it will be brushed back into the RAID, which can reduce the frequency of data.The checksum and requested operation is separated, in order to eliminate the I/Otime, which will improve the performance of the data. Through test analysis, cache policy for dealing with small data blocks can play a good role. It can solve the problem of poor performance of small data reasonably and efficiently.