Research On Optimization Technology Of Hybrid Storage System With Different Medium Solid State Drives

Flash-based Solid State Drive(SSD)has been widely used,but I/O bottlenecks are still severe,and due to the limited number of SSD erases and writes,lifespan has also become a prominent problem.The new storage device Optane Memory based on 3D XPoint technology has begun to apply,and with the advantages of latency,read performance,and life,it provides the possibility of speeding up the traditional SSD and extending the life of traditional SSD.However,the current hybrid storage system is not suitable for the hybrid devices with Optane Memory and SSD.If the scheduling strategy is not modified,the performance advantages of two disks cannot be fully utilized.In view of the difference in I/O performance,capacity,etc.between Optane Memory and the traditional SSD,the article implements a consistent access file system Hybrid F2 FS based on F2 FS for the hybrid storage system with Optane Memory and SSD,which optimizes read and write control and data migration to improve I/O performance.In view of the shortcomings of the existing hybrid storage system,such as high synchronization overhead and insufficient data attribute perception,Hybrid F2 FS implements hybrid device management at the file system level,and under the premise of ensuring consistency,the same storage space of Optane Memory is used for both cache and persistent storage,no data flushing required,reducing consistency synchronization overhead.According to the characteristics of data,the system stores metadata in Optane Memory to improve the retrieval speed,proposes a data scheduling strategy based on scoring,which combining a variety of factors to determine the location of data writing to reduce the amount of later data migration to improve system performance.In addition,Hybrid F2 FS monitors the data reading and writing process,recognizes hot and cold data,and flexibly migrates data between devices to further improve system performance.In order to alleviate the cache pollution caused by historical counting and take into account the characteristics of the device,an improved LFU algorithm based on two-level counter of time and I/O granularity is proposed to improve the recognition effect of cold and hot data.Aiming at the problem of segment space fragmentation caused by low segment effective data rate and the impact of data migration on user I/O,a data migration strategy based on segment effective data rate and migration window is designed,and different migration methods are adopted according to the effective data rate to reduce migration overhead.At the same time set the migration buffer to reduce the impact of the data migration process on user I/O.The experimental results show that,compared with the F2 FS using SSD,Hybrid F2 FS has significantly improved I/O performance,in which the access latency is reduced by 37.3%-80.1%,and the I/O throughput rate is increased by 59.8%-415.8%.Under the same test environment,compared with Bcache and Flashcache,the latency is reduced by 31.3%-86.8%,and the I/O throughput rate is increased by 47.2%-319.9%.