Re: Proposal: B_ARRIER (addresses wapbl performance?)

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

On Wed, Dec 10, 2008 at 08:16:45AM +1100, Daniel Carosone wrote:
> [...]
> Ordering constraints would produce a self-consistent journal. However,
> using ordered journal writes would also force other (unrelated)
> pending unordered writes out as well, adding latency to the
> journalling.  The journalling is already an overhead penalty (more
> total writes than would otherwise be required if dependencies could be
> fully ordered) and needs to jump the queue if it is to offer a latency
> reduction (reduced number of writes needed for *this* transaction to
> be acknowledged to upper layers).

But is there unordered writes, relative to the journal ? I guess there
are if there are different filesystems on the same storage. But,
unless I'm mistaken, all writes from the same filesystems should not be
reordered around journal writes (be it data or metadata writes), so either
you use tagged queuing or implement the write barriers in software with
the write-back cache disabled (which is what we're doing now). 

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--