On 20/04/07, Rick James <rjames@stripped> wrote:
Hi Rick,
I'll answer based on what I understood of your email, sorry if I got you wrong.
> Yeah, but what's the performance like when
> * the database is on the other side of the country?
It'll be lousy anyway, independent of RDF or not. This is the main
reason because Yahoo! can't use MySQL Cluster after all.
> * the data is too big to fit in RAM?
I'd rather assume that data will *always* be too big to fit in RAM and
use dynamic allocation. In this case the dynamic allocation should be
done on the other side (ie. restricting the dataset) and you'll only
get what you're looking for.
> We've been burned by simple-minded implementations of RDF that did not
> account for such.
True! I've seen and been talking to people using RDF and all of them
just gather all files possible, open them and load everything in
memory. Unfortunately my data set is far too big to fit on any
computer memory I have access.
This is why having a MySQL server on the other side and being able to
query and get only what I need would be a must. The added delay is
acceptable if the quality is good.
Before the internet, people used to have instant results to their
queries. the internet added a delay to see webpages, services,
documents and today it's already acceptable to have a "few seconds" of
delay for any page because after all you don't have all that
information at your desktop.
With a distributed data model it's the same thing. Today, you're used
to get the results of a query instantly because you're running MySQL
locally, but you can't store all the information in the world on your
server therefore the "added delay" will be acceptable if the
information you need is important.
It also gives you unmatched scalability and freedom from any company or vendor.
cheers,
--renato
Reclaim your digital rights, eliminate DRM, learn more at
http://www.defectivebydesign.org/what_is_drm
