Test Apache Commons IO ReadAheadInputStream performance

[lfcnassif]

Most of our current IO is synchronous. ReadAheadInputStream provides asynchronous reads, reading bytes ahead in a background thread, so they are already in memory when requested. I'm not sure if it would make processing noticeably faster, because while some threads are blocked on IO, usually others are using CPU. But sometimes I've seen some kind of CPU usage patterns like Hills, opposite to IO usage, when processing large files like unallocated, this situation may benefit of ReadAheadInputStream. Or maybe this would make processing slower because of more threads and context switches, just testing...

[lfcnassif]

Processed a case with 2.2M items and processing time increased in more than 50% (much more than I expected), so I'm closing this.