Re: Bio pool & scsi scatter gather pool usage

[Alan Cox]

> This is exactly the problem; I don't think it's going to be unusual to
> see volumes that have a variety of mappings.  For example the
> 'journal' area of the lv with a single fast pv, 'small file' area with
> a linear mapping across normal pv's, and finally a 'large file' area
> that has a few slower disks striped together.

Optimise for the sane cases. Remember the lvm can chain bio's trivially
itself. Its a lot cheaper to chain them than unchain them

> The last thing I want in this situation is to split up all the io into
> the lowest common chunk size, in this case the striped area which will
> typically be  < 64k.

The last thing I want is layers of bio support garbage slowing down a 
perfectly sane machine that does not need them...

Alan