> -----Original Message-----
> From: Greg Stark [mailto:gsstark@mit.edu]
> Sent: Thursday, April 14, 2005 12:55 PM
> To: matt@followers.net
> Cc: pgsql-performance@postgresql.org
> Subject: Re: [PERFORM] How to improve db performance with $7K?
>
> "Matthew Nuzum" <matt.followers@gmail.com> writes:
>
> > drive A has capacity C and spins at 15K rpms, and
> > drive B has capacity 2 x C and spins at 10K rpms and
> > all other features are the same, the price is the same and
> > C is enough disk space which would you choose?
>
> In this case you always choose the 15k RPM drive, at least
> for Postgres. The 15kRPM reduces the latency which improves
> performance when fsyncing transaction commits.
I think drive B is clearly the best choice. Matt said "all
other features are the same", including price. I take that to
mean that the seek time and throughput are also identical.
However, I think it's fairly clear that there is no such pair
of actual devices. If Matt really meant that they have the same
cache size, interface, etc, then I would agree with you. The
15k drive is likely to have the better seek time.
> The real question is whether you choose the single 15kRPM
> drive or additional drives at 10kRPM... Additional spindles
> would give a much bigger bandwidth improvement but questionable
> latency improvement.
Under the assumption that the seek times and throughput are
realistic rather than contrived as in the stated example, I would
say the 15k drive is the likely winner. It probably has the
better seek time, and it seems that latency is more important
than bandwidth for DB apps.
> > Would the increased data density of the higher capacity drive
> > be of greater benefit than the faster spindle speed of drive
> > A?
>
> actually a 2xC capacity drive probably just has twice as many
> platters which means it would perform identically to the C
> capacity drive. If it has denser platters that might improve
> performance slightly.
Well, according to the paper referenced by Richard, twice as many
platters means that it probably has slightly worse seek time
(because of the increased mass of the actuator/rw-head). Yet
another reason why the smaller drive might be preferable. Of
course, the data density is certainly a factor, as you say. But
since the drives are within a factor of 2, it seems likely that
real drives would have comparable densities.
__
David B. Held
Software Engineer/Array Services Group
200 14th Ave. East, Sartell, MN 56377
320.534.3637 320.253.7800 800.752.8129
