> Some people, including me, have problems with the nfs performance on a
> cluster with Linux PC clients and DEC Alpha OSF servers. I have tested
> the kernel patch:
> ftp://ftp.linux.org.uk/pub/linux/alan/2.2/patch-2.2.2-ac4.bz2 which
> gives a better performance (400 kB/s instead of 250 kB/s with the
> unpatched kernel), when i mount with options rsize=8192,
> wsize=8192. If i tried it with sizes of 16384 (perhaps a bad idea),
> the copy operation seams to use the full bandwidth (about 1MB/s), but
> the file on the server has a lenght of 0. Is it possible, to configure
> the server to manage that buffer sizes?
Hi,
The NFSv2 protocol specifies that the maximum allowed rsize/wsize value
is 8k. Thus, the immediate answer to your qustion is 'No'.
NFSv3 allows for larger values, and is the reason why the
maximum is higher in Alan's patches (NFS_MAX_FILE_IO_BUFFER_SIZE
should have been set to 8192 in nfs_fs.h for joe user).
Unfortunately, the full NFSv3 client is not yet ready; it is under
development, and will hopefully appear in Linux 2.3.
Cheers,
Trond
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/