ext2 large_file

[Bruce Evans]

On Tue, 1 Nov 2005, Ivan Voras wrote:

> On Tue, 1 Nov 2005, Bruce Evans wrote:
>
>> On Mon, 31 Oct 2005, Ivan Voras wrote:

> [for ext2fs under FreeBSD]
>>  first few cylinder groups full and the rest unused, where Linux would
>>  use all the groups fairly evenly.
>
> Not so good. I suppose this is not easaly fixable?

Not very easily, but not very uneasily either.  We can either obtain
a block allocator from NetBSD (I think it is similar to ffs's, but
specialized to ext2fs), or use the current Linux allocator (I think
we have an old version already -- ISTR a comment saying that it is
used for checking only).

>> - ext2fs is about twice as slow as the other 2 (worse for non-async 
>> writes).
>>  It is mostly because the block size is very small, and although this
>>  only necessarily costs extra CPU to do clustering, FreeBSD is optimized
>>  for ffs's default block size and does pessimal things with ext2fs's
>>  smaller sizes.
>
> These effects are also very noticable with NTFS (default block size=4096 for 
> all/most partition sizes) and FAT32 on smaller drives (where bs=4096 fits FAT 
> in 8MB).

4K isn't too bad, except possibly on arches with a page size of 8K -- see
my benchmark output for msdosfs.  Clustering is certainly required to
get good results for large files.

Bruce