Techniques for reducing long-term data movement on shingled magnetic recording drives.

摘要

Shingled Magnetic Recording (SMR) is a means of increasing the density of hard drives that brings a new set of challenges. SMR drives increase storage density by overlapping the write tracks (shingling) to be as wide as the read head. Due to the nature of SMR disks, updating in place is not an option. Holes left by invalidated data can only be filled if the entire band is reclaimed, and a poor band compaction algorithm could result in spending a lot of time moving blocks over the lifetime of the device. I propose using write frequency to separate blocks to reduce data movement and provide three band compaction algorithms that implement this heuristic in varying degrees: cold-weight, empty-separation, and separation+cold-weight. I demonstrate that using the techniques from the log-structured file system (LFS) perform worse than the greedy baseline algorithm. I demonstrate how my algorithms provides a reduction in long-term data movement. The cold-weight algorithm results in 1.81 times fewer blocks moved during required data movements when compared to naive approaches to band management. The empty-separation algorithm results in 135 times fewer blocks moved during required data movements compared to the greedy baseline algorithm. The separation+cold-weight algorithm results in moving up to 319 times fewer blocks during required data movements compared to the greedy baseline algorithm.