Focus is now on directory cache..
While current implementation can retrieve data for ~1000 directories per second (with -la), new implementation can retrieve them for nearly 100k directories per second (+ it gets permissions for any files in directory without additional cost) Amount of memory copying has been reduced to minimal by using "reference counters" for cached objects. Disk access is far more optimized.... and .ioFTPD files redesigned
I'll also add some statistic cookies (cache hits, misses, flushes, hash collisions)
|