We now present a case study that employs range writes to improve journal update performance. Specifically, we show how a journaling file system (Linux ext3 in this case) can readily use range writes to more flexibly choose where each log update should be placed on disk. By doing so, a journaling file system can avoid the rotations that occur when performing many synchronous writes to the journal and thus greatly improve performance.
Whereas the previous section employed simulation to study the benefits of range writes, we now utilize a prototype implementation. Doing so presents an innate problem: how do we experiment with range writes in a real system, when no disk (yet) supports range writes? To remedy this dilemma, we develop a software layer, Bark, that emulates a disk with range writes for this specific application. Our approach suggests a method to build acceptance of new technology: first via software prototype (to demonstrate potential) and later via actual hardware (to realize the full benefits).