Very Computer

    nfs write speed comparison, 100basetx, linux to
    freebsd/solaris, freebsd to solaris/freebsd,
    solaris to freebsd. (missing * to linux,
    solaris/solaris).

    basic conclusions: freebsd nfs performance ~= solaris
        performance.  linux performance 10x worse.

    (i'd appreciate recomendations on fixing the linux
    performance - one must internetwork).  i'd also be
    interested in other peoples nfs results.

            -elh

client:         twister. (red had linux 5, p2/233)
command:        time cat /boot/vmlinuz-2.0.32*4 > t.u   (1785124)

server:                                         write speed
           rsize/wsize                          (mbyte/sec)
hemlock v3 1024/1024:   1:27.39                 0.02            freebsd
2.2.6
xy      v3 1024/1024:   0:45.53                 0.04            freebsd
3.0-971225
stealth v3 1024/1024:   0:19.42                 0.09            solaris
2.5.5
hemlock v3 8192/8192:   0:14.00                 0.13
xy      v3 8192/8192:   0:08.50                 0.21
stealth v3 8192/8192:   0:05.81                 0.31

client:         xy.    (freebsd 3.0-971225-SNAP, ppro-150)
command:        time cat /kernel*8 > t.u                (11017089)

server:                                         write speed
                                                (mbyte/sec)
hemlock v3 1024/1024:   0:08.35                 1.32
hestia  v3 1024/1024:   0:08.72                 1.26
stealth v3 1024/1024:   0:05.04                 2.19
hemlock v3 8192/8192:   0:03.61                 3.05
hestia  v3 8192/8192:   0:03.48                 3.17
stealth v3 8192/8192:   0:01.73                 6.37

client:         stealth. (ultrasparc 1 150mhz, solaris 2.5.5)
command:        time cat /kernel/genunix*16 > t.u       (14196544)

server:                                         write speed
                                                (mbyte/sec)
hemlock v3 1024/1024:   0:05                    2.84
hestia  v3 1024/1024:   0:06                    2.37        freebsd
3.0-971225
hemlock v3 8192/8192:   0:03                    4.73
hestia  v3 8192/8192:   0:04                    3.55

In the meantime, you should be able to get somewhat better performance
by increasing the block size for read and write--i think 8192 is
recommended.  I don't have a linux box to check, but I think you want
mount options like rsize=8192,wsize=8192...

--sf

Mount options for nfs

        ...

       Especially useful options include

       rsize=8192,wsize=8192
              This will make your nfs connection much faster than
              with the default buffer size of 1024.

Changing the (r|w)size improved performance noticeably, but probably
stilll
quite slow.  Haven't tried the implementation in the 2.1 kernels though.

tom