Research Of Endurance-Aware Data Layout For SSD Storage Clusters

With the rapid development of science and technology,Solid State Drive(SSD)has been widely used in large-scale storage systems.However,SSD has a limited life.When the number of erasures reaches a certain level,performance and data reliability will decrease.Existing static data layout methods do not take into account the difference in wear between SSD devices.Existing dynamic data layout methods use data migration to balance wear between SSDs,adding additional write overhead to the SSD.Two data layout methods are proposed to solve the problems existing in the data layout methods for SSD storage clusters.One is the Endurance-Aware Static Data Layout(EASDL),which selects the initial position for the data.Reduce the degree of cluster wear imbalance by letting the most worn SSDs take on less data.The other is the EnduranceAware Dynamic Data Layout(EADDL),which balances the wear of the cluster through data migration.First,decide whether to perform data migration based on the difference in wear of the cluster.Secondly,based on the number of write of the SSD,the amount of migrated data is quantitatively calculated.Finally,considering the out-of-place update feature of SSD,the strategy of migration on update is used.Write directly to the migrated target device when there is an update request for the data to be migrated,thereby reducing unnecessary write operations during data migration.The experimental results show that the EASDL method can improve performance and reduce the wear difference of 19.4%~26.3% compared with the HASH method.Compared with the HASH method,the HASH+EADDL method can reduce the wear difference of 49.3%~71.1% and improve the performance by 2.1%~8.9%.Compared with HASH+SWANS method,HASH+EADDL method can reduce data migration by 24.7%~79.6%.The EASDL+EADDL method can reduce data migration by 47.9%~87.9% compared with the HASH+SWANS method,and can improve performance by 2.7%~11.9% compared with the HASH method.