> Hello,
>
> I have installed a kernel based on the cvs on this morning with gcc 2.95.3
> (20010219) and the loop fix patch applied (previously I was standing in
> /usr/src/linux instead of ~/cvs/linux-2.4-xfs that's why I could not apply it
> ).
> The loop fix fixed my crash 'xfs over loopback'. The performance for a
> PIII/666Mhz 320Mb RAM, 20Gb Seagate HDD:
> Bonnie
> -------Sequential Output-------- ---Sequential Input-- --Random
> --
> -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks-
> --
> Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %C
> PU
> 1* 900 8113 92.5 19879 33.1 4735 3.6 7818 83.2 21999 9.2 95.3 1
> .0
>
> Bonnie++
> Version 1.00h ------Sequential Output------ --Sequential Input- --Rando
> m-
> -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks
> --
> Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %
> CP
> esparrall 900M 7194 89 21480 31 4957 8 5250 58 27062 11 75.3
> 0
> ------Sequential Create------ --------Random Create------
> --
> -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete
> --
> files:max:min /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %
> CP
> esparrall 16 2974 76 +++++ +++ 3649 82 2755 72 +++++ +++ 1759
> 45
>
> Bye, GCS
Are these xfs over loop numbers? They are deceptive, since xfs thinks it is
writing to a device and when I/O completion comes back on a write it has
really gone to the device. In the loop case it has not gone to the device
it has gone to the buffers on the underlying file. For me xfs over loop in
an xfs filesystem is faster than the underlying xfs filesystem because
of this.
Steve
p.s. a fairly major change to the read path should arrive today, it should
not affect bonnie, but it will affect larger reads.
|