A Research Of HDD Caching System Based On SSD

With the advent of cloud computing and big data, a very high requirement of computer storage medium’s capacity and performance is put forward. The emergence of solid state disk meet this demand from a certain extent. But in practice, the contribution of performance made by solid state drive has not been released in full.Solid state disk(SSD), as a new kind of storage device, has been introduced into data centre because many of its advantages, such as compact size, low power consumption and high performance compared with traditional hard disk drive(HDD). But due to its relatively high price and low storage density, currently, it is a long run to completely replace hard disk drive with solid state disk. Actually, most of times, HDD and SSD exist in a system at the same time, and only a part of SSD’s space is used as storage. Thus, in author’s point of view, the IO performance of the storage system will be improved, if a software can configure part of the SSD as common storage while leaving unused part of SSD as a cache for HDD.In order to solve this SSD and HDD coexist but unreleased performance problem, this paper proposes a SSD based HDD caching system solution: using SSD as cache for HDD to improve HDD’s IO performance with multiple cache replacement algorithms.Based on algorithms depending on data locality theory, the frequently accessed hot data on HDD will be gradually copied to SSD, while leaving cold data on HDD. The latency will be very low to access hot data again. LRU, LFU and a new improved cache page replacement algorithm are implemented to distinguish hot and cold data. This new improved algorithm comprehensively considers time and frequency factors while its hit rate is relatively high and its time and space complexity is low. User can specify a SSD volume as cache’s storage and a HDD volume as a cached device. Software is implemented as a Windows®storage volume filter driver running in two modes: SSD as cache with write-hrough strategy and SSD as cache with write-back strategy. Mode differences are:the write-through strategy only cache read request, while write-back strategy cache both read and write request.To further improve performance optimization result, the cache system implements the support for big-cache-block: make the cache block size bigger than the aligned size ofa request. Which reduces the query times for the big request and improves efficiency of data copies.The proposed solution not only take performance advantage of SSD but also avoid costs to transfer all storage medium from HDD to SSD. With SSD cache, the I/O performance of HDD shows a 2-3x enhancement.