iJournaling: Fine-Grained Journaling for Improving the Latency of Fsync System Call

For data durability, many applications rely on synchronous operations such as an fsync() system call. However, latency-sensitive synchronous operations can be delayed under the compound transaction scheme of the current journaling technique. Because a compound transaction includes irrelevant data and metadata, as well as the data and metadata of fsynced file, the latency of an fsync call can be unexpectedly long. In this paper, we first analyze various factors that may delay an fsync operation, and propose a novel hybrid journaling technique, called ijournaling, which journals only the corresponding file-level transaction for an fsync call, while recording a normal journal transaction during periodic journaling. The file-level transaction journal has only the related metadata updates of the fsynced file. By removing several factors detrimental to fsync latency, the proposed technique can reduce the fsync latency, mitigate the interference between fsync-intensive threads, and provide high manycore scalability. Experiments using a smartphone and a desktop computer showed significant improvements in fsync latency through the use of ijournaling.