List:General Discussion« Previous MessageNext Message »
From:Benjamin Grosman Date:May 18 1999 4:29am
Subject:Re: Database distribution
View as plain text  
> Well, Oracle lets you use a big shared RAID array (or any kind of drive
> array, they could just be concatenated drives, as long as they're shared)
> to allow multiple database servers to serve requests from a single data
> source.  They've got a ton of cross-server locking work done to make that
> happen tho, and I don't think MySQL's gonna be doing that any time soon.

Distributed Database theory is an extremely difficult field. Although much
research has been done on it since the 70's, no real major breakthroughs
have yet been made, and there are really no commercial products available
to perform any real multidatabase operations.

Oracle's solution was at my last check not much more than a multiplexer;
it still relies on one machine to which everyone connects, which passes on
connections to other databases which it knows about. This box is also
responsible for intra-database deadlock detection/prevention, (2|3)PC, and
all standard operatins such as selects and joins, and optimization of
these. Many operations across multiple databases are extremely expensive,
and as such optimization is a very important field for distributed
databases.

However, judging by very recent releases, (i.e April 1999), there may be
more to the Oracle product now. 

A recent release from O'Reilly and Associates publishers entitled "Oracle
Distributed Systems" by Charles Dye may be of help for those interested in
this. However, it may still be the case that Oracle's Distributed Database
technology is still extremely formative - if any major breakthrough had
been made, then I am sure they would be making much more noise about it
than they seem to be.

More information on this book is available from:

http://www.amazon.com/exec/obidos/ASIN/1565924320/o/qid=927000625/sr=2-1/002-2180474-6801434

I don't have this book. I only just discovered it now. I am buying it
right now. Those who are interested, I will let know if it is any good.

That having been said, for those more interested in the science behind
Distributed Databases and why they are still such a difficult problem, and
therefore why seemingly so little progress has been made in the
commercialisation of such technology, the following may be of interest.

One (reasonably OK) introduction to the field is available in "Distributed
Database Systems" by David Bell and Jane Grimson.

More information on this book is available from:
http://www.amazon.com/exec/obidos/ASIN/0201544008/qid=927000625/sr=1-24/002-2180474-6801434

One item of note in this book is that in considering commercially
available systems, they dismiss Oracle's solution entirely and do not even
cover it, claiming it does not approach a level worth examining. However,
the book was published in 1992 - the authors' opinion may have changed
since then.

Another book (which is much better, much more rigorous, but also much more
mathematical in its treatment of the subject) is "Principles in
Distributed Database Systems" by M. Tamer Ozsu and Patrick Valduriez. I
recommend this book. Just make sure you get the second edition.

More information on this book is available from:
http://www.amazon.com/exec/obidos/ASIN/0136597076/o/qid=927000823/sr=2-1/002-2180474-6801434

(No, I'm not an amazon salesman, I actually *do* have these books, and
several others in the topic area).

Note, however, that this area is largely uncommercialised. Such solutions
as do exist are largely unsubstantial half-solutions, offering subsets of
true distributed functionality. It should therefore not be a surprise that
literature available on the topic is still highly ensconced in the arena
of theory rather than implementation, and frequently draws reference from
research papers and conference proceedings. 

Now, if any of you have entrepreneurial leanings, you've probably realised
that this is a potentially huge market area, ripe for explosive growth.
What you really need to do is to ask yourself why others haven't done it
first, and why a company like Oracle made so little progress until now,
and even now may still be only a subset of true functionality.

However, that isn't to say it wouldn't be nice if someone could come up
with a good workable solution to the problem. It would of course be even
nicer if it used MySQL and operated under the same generous licence
conditions as MySQL. (Many thanks to Monty and the gang at TCX).

Hope this helps,

Ben
---
perl -wle '$_ = 1; (1 x $_) !~ /^(11+)\1+$/ && print while $_++'

Thread
Database distributionDavid Johnson18 May
  • Re: Database distributionSasha Pachev18 May
    • Re: Database distributionunknown18 May
      • Re: Database distributionBenjamin Grosman18 May
        • Re: Database distributionMichael Widenius24 May
      • Re: Database distributionSasha Pachev24 May
RE: Database distributionPete Lancashire24 May
Re: Database distribution(David Sklar)24 May