On 4 oct, 18:14, Elcaro Nosille <Elcaro.Nosi... (AT) mailinator (DOT) com> wrote:
A total waste of time and money...(it does cost 67$ !!!)

Except from the first few lines...

*In a relational database, data is organized into table format*

On 4 oct, 19:05, Elcaro Nosille <Elcaro.Nosi... (AT) mailinator (DOT) com> wrote:
redefine RM.

http://www.amazon.com/dp/0201485559?...BE2TV51AMAPF4&


can smell cookbook declarations and non commitment to terminology from
miles.

Database Tuning by Dennis Shasha $ Philippe Bonnet (ISBN
1-55860-753-6) is a pretty good overview. But I would head for
academic publications like TODS from ACM and look at articles. Cheap,
fast storage and parallelism in the hardware is changing everything.

--CELKO--

--CELKO-- wrote:
hmmm. You've been following my sundry blogging?? Just kidding. But
not about the point. I don't follow hardware all that closely; then I
ran across a reference to an interview with Linus Torvalds, which I
read. In it, he referred to file systems and the looming presence of
solid state disks.

Which got me to thinking: if there's no rotational delay in the file
system, then what happens to the standard objection to the Relational
Database; joins are toooooo sloooow. And then, why not, finally, follow
the mantra: one fact, one place, one time?

Mayhaps this will shut up the knuckleheads.

I was looking at 8GB memory sticks at Costco yesterday under $20.00.
Look at WX 2 (nee White Cross), Teradata, SAND and Vertica for column
oriented RDBMS products.

In my next book, I carry things one step further. JOINs will become
cheaper than computations in the near future -- they are mostly
equality tests or simple comparisons. So more stuff will be done by
table look up in parallel hardware.