Re: Book about query processing / query optimization searched
Cimode
> I'm looking for a book that describes query-processing and
> query-optimization according to the current state of the art.
>
> Has anyone read the following title?
>
> "Principles of Database Query Processing
> for Advanced Applications"
> ISBN-10: 1558604340
> http://www.amazon.com/dp/1558604340/
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*
Cimode
> Cimode schrieb:
>
> > A total waste of time and money...(it does cost 67$ !!!)
>
Cause you will get a better value by people who do not continuously
redefine RM.
> > Except from the first few lines...
> > *In a relational database, data is organized into table format*
>
> Do you think you can estimate the quality of a book by evaluating
> the first lines?
I dunno but thanks to some people like F.PASCAL and a few others, I
can smell cookbook declarations and non commitment to terminology from
miles.
Cimode
> Elcaro Nosille wrote:
> > Cimode schrieb:
>
> >> 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*
>
> > the first lines?
>
> if the first line is silly, odds are you don't.
>
> As Groucho Marx said when asked to write a jacket blurb for a supposedly
> humourous book: (something like) "from the moment I picked it up until I
> put it down I couldn't stop laughing. Some day I'm going to read it".
Or when he simply asks a lady in a room with a cigar in his hand *Do
you mind if I don't smoke?*
--CELKO--
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--
MeBuggyYouJane
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.
--CELKO--
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.
Cimode
> --CELKO-- wrote:
<<I suppose column-based representations are superior when joins never
involve more than one per relation but the other complications will
attract implementers, as usual. >>
A key fundamental issue...
Question is: how could a column-based physical relation or relation
operation representation be superior in *any* case to some other
representation. Another way to ask the question: could join
consumption of IO resources under a column based system be increasing
any otherwise than exponentially as the number of join operation
increases?