SMR drive performance analysis under different workload environments

As the main stream of Hard Disk Drive (HDD) techniques, Shingled Magnetic Recording (SMR) drives have unique features different from conventional disk drives, e.g., append-only (sequential) write, indirect address mapping and garbage collection. Batch process is also designed to fully utilize the sequential write property to improve the SMR drive performance. The selection of different system parameters and policies affects the system performance and capacity efficiency. However, there is no dedicated analytical tool available so far to guide the parameter and policy selection. A queuing model is built and solved through a Markov chain process for the SMR drive to analyze the system performance under different kinds of system settings and workload environments. the control policies and parameter settings are also studied to explore their relation to the system performance of SMR disks, from the point of views of both analytical model and numerical simulation. An adaptive Garbage Collection (GC) policy is proposed to automatically select the foreground GC and background GC to provide consistent drive performance. A SMR drive simulator is further developed to validate the model and check the performance impacts of different drive parameters. We illustrate the similarity of the analytical and simulation results, and show that our tool can be utilized as the guide of the SMR drive design