In vdev_cache.c: 76 int zfs_vdev_cache_max = 1<<14; 77 int zfs_vdev_cache_size = 10ULL << 20; 78 int zfs_vdev_cache_bshift = 16; 79 80 #define VCBS (1 << zfs_vdev_cache_bshift)
So VCBS is 64k, but: 39 * the result. In the best case, this can turn 256 back-to-back 512-byte 40 * reads into a single 128k read followed by 255 cache hits; this reduces Says the readahead is 128k. IIRC, ZFS is documented to do 128k readahead in higher level documentation as well, but it looks like it's doing 64k readahead to me. -- This messages posted from opensolaris.org
