Font Size: a A A

Design And Implementation Of A Two-level Cache Algorithm Based On The RAID Storage System

Posted on:2016-08-04Degree:MasterType:Thesis
Country:ChinaCandidate:Y MenFull Text:PDF
GTID:2348330479953357Subject:Computer system architecture
Abstract/Summary:PDF Full Text Request
With the amount of global data growing explosively, the large storage system which is consturcted by traditional RAID(Redundant Array of Independent Disks) has been unable to meet the needs of the application in performance. SSD(Solid-state drive) is a new kind of storage medium which has high performance and low power cost. It can be introduced into the traditional storage system as a cache to effectively improve the performance bottleneck. This is a hot topic in the current storage areas and the study of multi-level cache management and data management in the hybrid storage system is important.Introduce SSD as a second cache into the traditional RAID and achieve two-level cache architecture. According to the different characteristics between RAID and SSD, propose a region dynamical adjusting algorithm and DRAM is adaptively divided into SSD buffer and RAID buffer. SSD buffer uses a replacement algorithm which is based on the caching block's cost of victim. The algorithm evaluates the cost of victim recording to the heat of cache block and read/write speed of SSD, and replaces the block with least cost first. RAID buffer use the LRU algorithm which is based on stripe. The algorithm can effectively utilize the parallel I/O of HDDs and improve the read/write throughput rate of the RAID. At the same time, it can reduce the write penalty in RAID which has parity informations. A cache bock which is eliminated from RAID buffer will go through a hot data filter, the hot data will be stored to the SSD and reduce the read/write to the RAID. SSD use a modified set-associative mapping which can speed up the performance of writing data to RAID.Implement these algorithms and ues database testing tool to do a series of tests. The results show that the TPS(Transaction Per Second) of the storage system which uses the SSD can improve 459.2%; Compared with the fixed region algorithms, the TPS of the region dynamical adjusting algorithm can improve 7.9%. Furthermore, Compareing with the CFLRU algorithm and the LRU-WSR algorithm, the algorithm which is base on the cost of victim works well at speeding up the performance of system and reducing the write operations to SSD.
Keywords/Search Tags:Solid-state drive, Dynamical adjusting algorithm, Cost of victim, Hot data
PDF Full Text Request
Related items