A segmented garbage collection method for solid-state storage devices

Abstract

The invention discloses a segmented junk recovery method for a solid-state storage device. The method comprises the steps of adding a cache which difficultly loses data after power failure and is specially used for recovering service junk and migrating effective data, and temporarily storing the effective data needed to be migrated by utilizing the cache in effective data migration of junk recovery operation; when erasure operation is performed, in a way of setting a marker for marking a plane or a die in which a target block subjected to block erasure operation is located, feeding back junk recovery operation completion to an upper layer and continuing to make a response to a coming request, namely, not making a response to the request by the plane or the die, thereby avoiding a time delay of the response caused by the fact that a write request allocates idle pages of the plane or the die; and at gap time when read and write requests come, writing back the effective data migrated to the cache which difficultly loses the data after power failure to the solid-state storage device. According to the method, the time delays of read and write responses caused by junk recovery can be effectively shortened without influencing the junk recovery efficiency.

Description

technical field [0001] The invention belongs to the technical field of garbage collection of solid-state disks, and in particular relates to a segmented garbage collection method of a solid storage device, which is used for reducing read and write time delays caused by garbage collection operations. Background technique [0002] In recent years, NAND Flash-based solid state drives (Solid State Drive, SSD) have been widely used due to many advantages such as high performance, low power consumption, high reliability, good shock resistance, and no noise. However, after the SSD is used for a period of time, its access performance will drop significantly. Garbage collection has a significant impact on SSD performance. Through testing on Intel SSD DCP3700 (800G), it is found that for random mixed requests with a read / write granularity of 4KB and a read / write ratio of 7:3. When the IntelSSD is an empty disk, its IOPS can reach 200,000, and its read-write bandwidth can reach 800MB / ...

AI Technical Summary

Problems solved by technology

This solution can solve the conflict between garbage collection and normal requests to a certain extent, but its interruption time selection is limited, and it can only occur during data migration, without considering the longest time-consuming erasure operation and reading and writing in the garbage collection process. request conflict

Images