[rabbitmq-discuss] New Persister

[Matthew Sackman]

Hi Jeremy,

On Sun, Mar 21, 2010 at 08:51:19PM -0500, Jeremy Dunck wrote:
> On Sun, Mar 21, 2010 at 5:58 PM, Matthew Sackman <matthew at lshift.net> wrote:
> ...
> > If you do choose to use the new persister, try to ensure the following:
> > 1) Try to make sure nothing else is writing to the disk that Rabbit is
> > using - Rabbit is pretty good at managing the position of the disk head,
> > but that tends to go wrong if other applications are writing at the same
> > time.
> 
> Sorry, this sort of confuses me.  I are you saying that in order to
> get best performance, Rabbit should be used with no apps using the
> same disk?  Or are you saying app-level bugs might occur if Rabbit is
> used with other apps using the same disk?
> 
> If you're just saying that sequential IO is faster and Rabbit tries to
> minimized seeks - sure, OK.  If it's the latter -- that you're
> literally managing the drive head -- can you elaborate on that
> decision?

It's the former. We mainly optimise by making sure that writes are
always append (so you should be able to get the full bandwidth of your
hard disc - fsync's permitting). The disk access patterns *should* be
sufficiently straight forward enough that OS disk prefetching and
caching works very well for reads.

Matthew