A Fast and Slippery Slope for File Systems

There is a vast number and variety of file systems cur-rently available, each optimizing for an ever growing number of storage devices and workloads. Users have an unprece-dented, and somewhat overwhelming, number of data man-agement options. At the same time, the fastest storage de-vices are only getting faster, and it is unclear on how well the existing file systems will adapt. Using emulation tech-niques, we evaluate five popular Linux file systems across a range of storage device latencies typical to low-end hard drives, latest high-performance persistent memory block de-vices, and in between. Our findings are often surprising. De-pending on the workload, we find that some file systems can clearly scale with faster storage devices much better than others. Further, as storage device latency decreases, we find unexpected performance inversions across file systems. Finally, file system scalability in the higher device latency range is not representative of scalability in the lower, sub-millisecond, latency range. We then focus on Nilfs2 as an especially alarming example of an unexpectedly poor scala-bility and present detailed instructions for identifying bottle-necks in the I/O stack