> On Wed, 20 May 1998, Sieger Ralf AI wrote:
>
> > I currently developing a fs which supports among other things
> > fragmentation. This means for better throughput I want to use the highest
> > possible blocksize. So if someone has an solution to this limit I would
> > cheer.
>
> It's simple. Use 4kB blocks for I/O, and use multiple blocks
> for on-disk allocation.
> ie:
> - you do I/O in multiple 4kB chunks
> - you allocate disk space per 256 4kB chunks
>
> This way, you have 1MB disk blocks. And if you use the
> ext2fs preallocation scheme, you can even free the 4k
> blocks which aren't used by the file when the file is
> closed...
What do you do with the freed 4k blocks? Novell has one answer --
somehow they "suballocate" this unused space, packing several small files
into a single block. When free blocks run short, though, the machine
goes into a tailspin looking for spare sub-blocks and does very little
else. (Axe one big file and your responsiveness comes right back...once
your delete request get serviced.)
-- Mark H. Wood, Lead System Programmer mwood@IUPUI.Edu Nothing is so dangerous as being too modern. One is apt to grow old-fashioned quite suddenly. -- Oscar Wilde- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu