Research And Implement On Three-level Cache In RAID

The RAID does not always provide us with satisfied performance at any environment.some dissatisfied facts about RAID was: The access of the alone disk in RAID are same as the common disk. And the slowly mechanical movement is still the limitation of RAID. So the performance of RAID can't improve more and the access time is still in ms. along with the expand of computer,the service of master storage,response in millisecond have not fit the needs of today;we can improve throught-out and transfers by parallel accessing data and parallel transferring data in RAID which have multi-disks work at one time.make use of these characteristic of RAID,we can improve the response time of storage system.but there are many requests(eg:small block random-access request in light load)care the response time of single request which was decided by the disk's access time and has a little of effect on throught-out and transfers.Thus now design a new Cache system on RAID,and the Cache system in RAID adopt three main strategy :The Cache system on RAID adopt prepay and dynamic allocate in global config the Cache system,and building the allocate model.all above make certain the optimal proportion of prepay and dynamic allocate theoretically;There are two level read cache in the Cache system on RAID.the read cache of level1 achieve temporal locality and the read cache of level2 achieve spatial locality;The design bring forward timing-move strategy and replace-move strategy. namely read cache of level2 move the accessed data recently to read cache of level2,and replace the data of read cache of level2 when the numbers of data blocks exceed a valve.and the timing-move strategy was proved successly implement the locally of time and improve the performance of Cache system by the emulational test.