Relational Versus Multi-Dimensional Database formats 
Author Message
 Relational Versus Multi-Dimensional Database formats

How can OLAP even be applied to a relational database format??  I see
that IBM is attempting to bridge their A/S 400 into a client/server
environment, taking advantage of the blah blah blah...

When will the relational database followers of the world recognize that
there is no competition between their format (and their SQL-based
front-end) and a true multi-dimensional database (especially one with a
good GUI front-end) when it comes to OLAP?  If we think about our
customers (rule 1!), they don't want data in record by record format,
they want it summarized!!

Now somebody's going to tell me that their Relational Database
processor/algorithm rivals processing speeds of the true OLAP
Multi-dimensional stuff....



Mon, 30 Mar 1998 03:00:00 GMT
 Relational Versus Multi-Dimensional Database formats


writes:
Quote:

>How can OLAP even be applied to a relational database format??

   snip

Quote:
>When will the relational database followers of the world recognize that
>there is no competition between their format (and their SQL-based
>front-end) and a true multi-dimensional database (especially one with
a
>good GUI front-end) when it comes to OLAP?

I have an article coming out in the October 30th issue of Information
Week on this topic, why not have a look and let me know what you think?
The gist of it is that there are approriate uses of both technologies,
but good evaluation may be obsolete in six months.

Roughly, here are recommendations:

Use MDDB for: bounded dimensionality, cross-dimensional calculations,
row-level calculations, read-write applications, rules-rich
applications and data marts

Use Relational/OLAP for: Fluid dimensionality, multidimensional view of
a data warehouse, very large scale (for now), data-rich applications,
rapidly changing dimensions.

We had a long-running thread on this topic a while ago, but it was
largely focused on speed and functionality. Actually, this is a gripe I
have about this group, that the discussion is a mile wide and an inch
deep, far too concerned with taxonomy and technical minutiae. That's
why we keep getting stuck on issues like this. The really interesting
issues in decision support are content and effectiveness, which are
never addressed in this forum.

--
_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/

_/
_/ Archer Decision Sciences, Inc.
_/ 133 E. De La Guerra Street
_/ Santa Barbara, CA 93101
_/
_/ Reengineering, DSS Architecture, Data Warehousing
_/ CA:  805.564.8672      NY:  914.987.9528    
_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/



Wed, 01 Apr 1998 03:00:00 GMT
 Relational Versus Multi-Dimensional Database formats
I am right with you on the real issues of DSS!  I replied to the
relational stuff because I have some gripes about traditional-thinking IS
folks beating up on MDDB, so I over-reacted the other direction...

I agree that the real challenge is not to calculate the structures or
find new ways to code screens, but instead it is to frame the front-end
to solve the business problem/problems.  Now, while we cannot standardize
business problems, (ie it would not make sense to attempt to generalize
on these sorts of issues), I do believe a viable discussion can be
maintained on methods of getting to the information systems needs.

In other words, do you have any ideas for methods of "translating"
business rules from the mouths of the decision-makers to the screen
itself?

Thats my interest!!

Mark

PS - Okay, okay, relational databases are a "must" for any record by
record retrieval, as well as for supporting requests that encompass too
many dimensions, or variables, to be effectively/efficiently handled by
an MDDB.  So they do have a place in this world :)



Wed, 01 Apr 1998 03:00:00 GMT
 Relational Versus Multi-Dimensional Database formats

says...
Quote:

>How can OLAP even be applied to a relational database format??  I see
>that IBM is attempting to bridge their A/S 400 into a client/server
>environment, taking advantage of the blah blah blah...

>When will the relational database followers of the world recognize
that
>there is no competition between their format (and their SQL-based
>front-end) and a true multi-dimensional database (especially one with
a
>good GUI front-end) when it comes to OLAP?  If we think about our
>customers (rule 1!), they don't want data in record by record format,
>they want it summarized!!

>Now somebody's going to tell me that their Relational Database
>processor/algorithm rivals processing speeds of the true OLAP
>Multi-dimensional stuff....

I see users every day who want to work simultaneously with summarized
and detailed data.  Identify the hot products by geography and then
give me names and addresses to market the hot ones to.  Find the
dimensional characteristics that indicate a high likelihood of fraud
and then give me the detailed records that fit that profile.  When are
MDDBMS zealots going to see that the summary stuff that their software
excels at (I know, I launched an MDDBMS) is only one part of the range
of business needs for data consumers.  An important part to be sure,
but only one part.

In fact, I understand that the vendors are not, by and large, caught in
that trap but are pursuing ways to integrate MD and RDB structures, at
both the application level and the software platform level.



Thu, 02 Apr 1998 03:00:00 GMT
 Relational Versus Multi-Dimensional Database formats
Quote:


>says...

>>How can OLAP even be applied to a relational database format??  I see
>>that IBM is attempting to bridge their A/S 400 into a client/server
>>environment, taking advantage of the blah blah blah...

>>When will the relational database followers of the world recognize
>that
>>there is no competition between their format (and their SQL-based
>>front-end) and a true multi-dimensional database (especially one with
>a
>>good GUI front-end) when it comes to OLAP?  If we think about our
>>customers (rule 1!), they don't want data in record by record format,
>>they want it summarized!!

>>Now somebody's going to tell me that their Relational Database
>>processor/algorithm rivals processing speeds of the true OLAP
>>Multi-dimensional stuff....

>I see users every day who want to work simultaneously with summarized
>and detailed data.  Identify the hot products by geography and then
>give me names and addresses to market the hot ones to.  Find the
>dimensional characteristics that indicate a high likelihood of fraud
>and then give me the detailed records that fit that profile.  When are
>MDDBMS zealots going to see that the summary stuff that their software
>excels at (I know, I launched an MDDBMS) is only one part of the range
>of business needs for data consumers.  An important part to be sure,
>but only one part.

>In fact, I understand that the vendors are not, by and large, caught in
>that trap but are pursuing ways to integrate MD and RDB structures, at
>both the application level and the software platform level.

Just a quick reply, since apparently this was covered previously (sorry,
I am new to the group:)).  What I am talking about is not just "business
needs for data consumers".  I'm talking about OLAP.  As far as I'm
concerned, OLAP means high level summary information, not relational
record by record information.  Business executives, a group I am
intimately familiar with, are not interested in "Sorry sir, I gotta do
this SQL thing for your what-if questions, but I'll get right back to
you."  They want summary info now, Now, for the un-initiated, means at
the moment, at their fingertips, real easy to use.  Show me the
relational product that does that, and I'll be amazed! (I'll also be an
investor!  MDDB's are a total pain to set up on the back end!  If there
are any serious relational summary/front-end tools, please respond!  I
would welcome the relief!)


Mon, 06 Apr 1998 03:00:00 GMT
 Relational Versus Multi-Dimensional Database formats


..
(previous author unrecorded, but whose opinions I agree with):
..

Quote:
>>When are
>>MDDBMS zealots going to see that the summary stuff that their
>>software excels at (I know, I launched an MDDBMS) is only one part of
>>the range of business needs for data consumers.  An important part to
>>be sure, but only one part.

>>In fact, I understand that the vendors are not, by and large, caught
>>in that trap but are pursuing ways to integrate MD and RDB
>>structures, at
>>both the application level and the software platform level.

>Just a quick reply, since apparently this was covered previously
>(sorry, I am new to the group:)).  What I am talking about is not just
>"business needs for data consumers".  I'm talking about OLAP.  As far
>as I'm concerned, OLAP means high level summary information, not
>relational record by record information.  

no, that isn't quite what you're talking about either.  you're talking
about storing data in pre-aggregated form vs storing at a detail level,
and you seem to be implying that pre-aggregated data cannot be stored
in a relational database.  you're not the first one to make that
mistake... but I'm beginning to think that there may be an evangelical
church of OLAPtology that is spreading this misinformation
systematically!

Quote:
>Business executives, a group I am
>intimately familiar with, are not interested in "Sorry sir, I gotta do
>this SQL thing for your what-if questions, but I'll get right back to
>you."  They want summary info now, Now, for the un-initiated, means at
>the moment, at their fingertips, real easy to use.  

for future reference, mark, most of us who frequent c.d.o. have been
initiated, but we don't all belong to the same fraternity.

Quote:
>Show me the
>relational product that does that, and I'll be amazed! (I'll also be
>an investor!  MDDB's are a total pain to set up on the back end!  If
>there are any serious relational summary/front-end tools, please
>respond!  I would welcome the relief!)

You seem to be making three implications here: first that relational
products are multidimensional, second, that this would be bad
thing/"total pain", and third, that OLAP tools are not multidimensional
databases.  Did I misinterpret what you wrote??

as a final note, you won't find many staunch defenders here of SQL as
the query language of choice for performing typical OLAP-type
operations.  I think one of the true breakthroughs in this area is the
recognition that hierarchies of business information can be treated as
a linearized "dimension", and that by limiting a tool to requesting
intersections of subsets of these hierarchical dimensions you can
greatly simplify the UI and get IS (and SQL) out of the loop.

   /jim



Wed, 08 Apr 1998 03:00:00 GMT
 
 [ 6 post ] 

 Relevant Pages 

1. Multi-dimensional vs Relational

2. Multidimensional versus Relational

3. Object databases versus relational databases

4. Looking for a multi-dimensional database

5. >>PW: OLAP Consultant - Oracle, Designer/2000, Discoverer, Impromptu, multi-dimensional database - JL (PW124989-10005PW)

6. Disconected Multi Dimensional Database?

7. multi-dimensional database

8. What is a multi-dimensional database?

9. theory of multi-dimensional database ?

10. Data warehouse & multi-dimensional databases

11. Info on Multi-Dimensional Databases

12. Disconnected Multi Dimensional Database?


 
Powered by phpBB® Forum Software