Research On Content-aware Garbage Collection Technology In Flash-based Storage System

Compared with the traditional disk media,the existing flash storage device has obvious advantages,such as high speed,low power consumption and small size.However,there is an obvious characteristic,that is Garbage Collection(GC),The GC operation causes Write-Cliff and Write-Amplification problems,which directly affects the performance and reliability of flash storage.On the other hand,data deduplication can significantly save the storage space,which has been widely concerned by industry and academia.Data deduplication eliminates the deduplicated data through hash verification,which can alleviate the effect of Write-Cliff and Write amplification caused by garbage collection to some extent.However,with the development of NVMe and 3D XPoint,the performance of flash storage rapidly improves.On the contrary,the online deduplication technology will affect the performance of the flash storage system.The thesis combines the data deduplication technology and garbage collection technology and proposes a content-aware garbage collection technology in the flash-based storage system,called DAGC.Traditional deduplication-based flash storage system performs online deduplication when data is written,so the hash computation delay will become the performance bottleneck of high-performance flash memory.The DAGC scheme does not deduplicate the data during the writing process,but deduplicates the duplicated data when the flash memory is performing the garbage collection operations.By taking the advantage of long-time garbage collection and the migration of valid data,DAGC carries out hash calculation so as to smartly conceal the calculation time of hash.At the same time,it also marks the data blocks according to the data reference count to store the data blocks with high and low reference counts separately.DAGC reduces the amount of data migration when performing garbage collection and improves the efficiency of the garbage collection process.Moreover,it significantly alleviates the Write-Cliff and Write-Amplification issues,thus improving the performance and reliability of flash storage system.The experimental results show that DAGC system significantly reduces the number of block erasure and migration pages during garbage collection.