Re: 5.x filesystem performance regression

[Edgar Fuß <ef%math.uni-bonn.de@localhost>]

> One suggestion would be to use time -c which should show disk I/O counts
> and see if there's any significant difference.

OK, but I'll not be doing that for all the parameter combinations:

5.1 WAPL, 32k RAID 5:
1. 2.2u 9.7s 6:17.70 3.1% 0+0k 595+17416io 9pf+0w
2. 2.1u 9.4s 3:13.03 5.9% 0+0k 0+24348io 0pf+0w

5.1 softdep, 32k RAID 5:
1. 2.1u 9.1s 6:04.22 3.1% 0+0k 597+11895io 0pf+0w
2. 2.0u 9.9s 5:02.16 3.9% 0+0k 1+11809io 0pf+0w

4.0.1 softdep, 32k RAID 5:
1. 2.0u 8.7s 2:13.29 8.1% 0+0k 6+11534io 10pf+0w
2. 2.1u 8.7s 0:19.92 54.4% 0+0k 17+11251io 0pf+0w

4.0.1 softdep, 16k RAID 1:
1. 2.1u 7.8s 0:18:99 52.8% 0+0k 565+11380io 0pf+0w
2. 1.9u 7.9s 0:12:24 81.0% 0+0u 16+10877io 0pf+0w

So I'd say there's no no significant difference in I/O counts.