CLDM: A Cache Cleaning Algorithm for Host Aware SMR Drives

Host aware SMR (HA-SMR) drives can effectively increase the capacity of hard disk drives. However, the cache cleaning algorithms implemented in the HA-SMR drives need to be improved. Current cache cleaning algorithms do not consider the characteristics of applications and usually bring too much data migration. In this paper, we propose a new cache cleaning algorithm called CLDM, which takes the characteristics of applications into account. It uses the “zone heat” to reflect the access frequency in the disk cache of a zone, and the “zone data migration” to reflect the data migration of a zone when cache cleaning is performed on the zone. When CLDM is performed, it first computes the “zone heat” for each zone which is currently buffered in the disk cache, and then computes the “average zone heat” for all the buffered zones. After that, CLDM computes the “zone data migration” for each buffered zone, and sorts all the buffered zones in the ascending order of their “zone data migration”s. CLDM first cleans the zones which satisfy the condition “the zone heat of a zone is less than the average zone heat”. And then it cleans the zones with less “zone data migration”s. Experimental results show that CLDM can effectively reduce the amount of migrated data during both the cache cleaning process and garbage collection process, and improve the performance of HA-SMR drives.