Implementation

In order to gain a more complete insight into how our prefetching methods would interact with a typical I/O system, we implemented predictive prefetching in the Linux kernel. Our implementation consisted of adding two components to the VFS layer of the Linux kernel, a model and a prefetching engine. Our models tracked file access patterns and produced a set of predictions for upcoming accesses. Our prefetch engine selected predicted files and prefetched their data into the page cache. The implementation itself consisted of less than 2000 lines of C code.