Hi, I tagged a new release today, since there were a few unfortunate issues with 1.25. The major issue was the throughput reporting being off by 1024/1000, which was very unfortunate. It would present higher throughputs than was really generated. Terse outputs were not affected, but the normal outputs were. From that output, the bw= reported for the specific job was off, as well as the aggrb=, minb=, maxb= in the group status report. All other numbers were correct. Another important fix is the one from Carl Henrik Lunde, where an off-by-one would cause fio to crash with multiple files. Likewise the fix from Radha Ramachandran fixes cases where we generated offsets that were too large, causing either early exit of fio or IO verification failures (if run with verify enabled). On the feature side, the bssplit= option now accepts different weights for reads and writes. The format is identical to previously, but you can now separate the read and write side with a ",". As an example, this: bssplit=4k/90:8k/10,4k/20:8k/80 would generate 90% 4k reads and 10% 8k reads on the read side, while generating 20% 4k and 80% 8k writes on the write side. I also added a simple job file to get some quick performance numbers of out SSD drives. It's basically just a test of sequential read and write and random read and write with 4k block sizes, to enable a quick check of whether the drive is crap or not. The 4k rand write test is the most interesting one, the others are mainly just sanity checks. Grab the new release from: http://brick.kernel.dk/snaps/fio-1.26.tar.bz2 or with git from: git://git.kernel.dk/fio.git or just git pull from your repo, if you are already using git. -- Jens Axboe -- To unsubscribe from this list: send the line "unsubscribe fio" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
