On Thu, Oct 27, 2011 at 9:56 PM, trskopo <trskopo@...> wrote:

>
> I just doing a little experiment to find out database performance on RAM vs
> on Disk. I use WD SATA 2 Green, 64MB cache for testing. My RAM speed is
> about 35x faster than disk. I have 2 databases (size is about 75 MB), one
> placed on a RAM and the other on disk, then I run select sql for both
> databases.
>
> SQL Statement finished about 47 seconds, on RAM and on disk. So there is no
> significant performance between RAM and disk. Performance on RAM shows when
> doing backup or restore routine.
>
> So I wonder, could it be that there is no need for faster hardisk when
> using firebird?

Counting on my (wrinkled and lizard-like) fingers, I just realized I've been
looking at database performance for about 35 years and I have no more clue
about how to measure it than I did when I started. In fact, when I started,
I thought I knew something, and now I'm pretty sure I don't. Net backward,
35 years. Sigh.

What counts is whether a specific database produces and saves data fast
enough for your application and users. And the mix of updates and searches,
and the locality of data meaning the cache hit rate. Perhaps for your
application, disk speed doesn't matter. I can construct a bunch of
situations where it does.

The real challenge is understanding where your application requires
resources and finding a database that puts its resources where you need
them.

Good luck,

Ann

>

[Non-text portions of this message have been removed]