Block Sizes

[Greg Lindahl]

On Thu, Oct 31, 2002 at 11:36:54AM -0500, Robert G. Brown wrote:

> You have at least a limited ability to tune filesystem block sizes and
> other parameters for good high bandwidth performance at the kernel
> level, before fortran or anything else sees it.  Read:
> 
>   man hdparm
>   man tune2fs
>   man mke2fs
> 
> I think you'll want to create your partitions with -T largefile4, 4K
> blocksize (the maximum, which I recall is also the kernel page size
> which is probably WHY it is the maximum) which allocates one inode for
> every 4 MB of disk.

The question was about making I/O faster. What you suggest wastes less
disk space on inodes, but won't speed up anything because ext2's usual
blocksize is 4K. Larger blocks doesn't necessarily improve things, it
really depends on whether their I/O is linear or random or whatever.

BTW, 4K is not the pagesize on all versions of Linux; it is the
pagesize for i386. Itanium, Alpha, sparc, sparc64, and probably others
have larger pagesizes.

> There are also wsize,rsizes that can be adjusted in nfs/fstab that
> adjust how data is buffered and written back to NFS mounts.

This is good advice; raising these to 32k often improves performance.
However, some older Linux kernels won't raise them above 8k.

Pedantically yours,

-- greg