Re: One Ring to Bind Them

"Anthony W. Youngman" <wol_at_thewolery.demon.co.uk> wrote in message news:DWM+oVL8Ny7AFwjT_at_thewolery.demon.co.uk...
> In message <CLgGc.26848$a24.25938_at_attbi_s03>, Marshall Spight
> <mspight_at_dnai.com> writes
[snip]
> >Ugh. Let's please not talk about disk reads.
> >
> So you're quite happy to give your clients a system that, running on a
> Cray, still makes an old Z80-based system look like a speed demon?
>
I'll jump in here though I've missed the meat of this discussion thread and I'm sure what I'm about to say has been discussed at length. Just let me know. I'll try to get caught up in the meantime.

> The whole reason we hammer on about disk accesses is because we KNOW we
> can't be beaten.

This is likely debateable. If the query and data access conforms to the physical organization of MV data, then this is probably true. When the query is ad-hoc and forces a full scan multiple times across multiple branches of a hierarchy, it is probably false. Plus the caching mechanisms of relational databases these days are quite advanced. If MV has to access disk more than a relational db, the question of who is faster is moot. However, if MV has as sophisticated caching and predictive algorithms, then we're back to the base assertion.

Plus, in ideal conditions, I betcha' an IMS system might be just as fast, if not faster. So, the assertion could probably be qualified as: "we KNOW we can't be beaten by a relational system."

[snip]

>
> Cheers,
> Wol
> --

[snip]

• Dan