Research On The Optimization Method Of Copyback-based Garbage Collection And Wear Leveling For NAND Flash

With the rapid development of information technologies such as mobile Internet,Internet of things,and cloud computing,the amount of data is exploding.The diversified big data applications put forward higher requirements for the performance and reliability of the storage system.Due to their high performance,high reliability and low energy consumption,NAND flash-based solid state drives(SSDs)have been widely used in embedded devices,personal computers,enterprise storage systems and data centers,and they have become the mainstream storage device.How to reduce the negative effect of the unique characteristics of NAND flash(such as wear write and out-of-place update)on SSD performance and reliability has become a hot topic.Garbage collection is an algorithm for defragmenting dirty blocks of flash memory,which has a direct impact on the performance and lifetime of flash memory.In traditional copyback-based garbage collection,the copyback feasibility detection(whether copyback can be executed reliably or not)leads to an expensive cost,significantly reducing the efficacy of garbage collection.To solve this problem,a fast garbage collection scheme based on copyback,called FastGC,is proposed.The quantitative correlation between the number of copyback execution and data failure is established at first.Based on the quantitative correlation,an efficient copyback feasibility detection scheme is proposed.Instead of transferring user data to the controller,the count of copyback execution is selected as the indicator to detect the copyback feasibility,which efficiently degrades the data transmission latency.Experiment results show that FastGC improves the read response time and write response time by up to 66.3% and 44.2% respectively,compared to the traditional copyback-based garbage collection.Wear leveling is used to balance the wear difference of flash memory blocks,which has a direct impact on the lifetime of flash memory.Copyback execution can lead to the inter-plane unbalance wear distribution and reduce SSD lifetime.To solve this problem and optimize FastGC,an efficient wear leveling algorithm based on FastGC,called FastGC-WL,is proposed.Motivated by the relationship between flash wear difference and the amount of data migrated per garbage collection,FastGC-WL evens out the amount of data migrated per garbage collection to eliminate the wear disequilibrium across planes.FastGC-WL employs an effective data migration leveler taking the data coldness and the erase count per plane into consideration to guide data migration.Experiment results show that FastGC-WL can efficiently reduce the wear difference of flash memory,compared to other wear leveling algorithms.