Re: [fuse-devel] large_write like large_read (again)

[Howie <how...@gm...>]

>I'm having the same problem with a small file system I'm working on.
>The lost in performance for file writing is about 30% and i think it's
>caused for the 4K blocks....

I 'm also have this problem .
This is the performance testing which use "Bonnie tools"
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
                  -------Sequential Output---------------------------=20
----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=20
local      100  28045  99.7 195131 101.0 321515 100.5 38494 101.1
1217829 95.1 108324.8
fuse1     100  24253  66.6   85965  21.0   65853   18.0 32531  82.6 =20
148139 14.5   15607.1
fuse2     100  29626  72.3   94936   22.3 123942   29.0 43572  99.6=20
1481160 100   39950.1
---------------------------------------------------------------------------=
--------------------------------------------------------

fuse1 without large_read ,fuse2 use large_read kernel_cache
when we use large_read the read performance is as good as local operation.
But  write will loss some performance.

Is it possible have large_write option ,the same as large_read.

howie