A large corporate bank has asked an independent  data warehousing
consultant to advise it about its IT systems.  The bank?s own IT
manager has recommended keeping the existing database system, which
manages the day-to-day bank transactions, and using an OLAP (OnLine
Analytical Processing) system to gain business intelligence.  The data
warehousing consultant disagrees, stating that a data warehouse (DW)
should be built for business intelligence, and a normal DBMS (DataBase
Management System) should be used for the day-to-day transactions. 
Critically compare and contrast these two suggestions, covering both
technical and business issues.

Hi

Have been given the above question and am confused as to what it is
asking for???....my opinion is that an olap should be used in
conjunction with a data warehouse???

Any feedback much appreciated

Regards

mandeep

---

Clarification of Question by thecanj-ga on 04 Jun 2006 07:58 PDT

Ok from what i understand wouldn't the consultant's idea be a more
effective approach as implementing only a DBMS and OLAP system would
mean the a olap query may suck resources from the database to such an
extent which may hurt transaction processing.

Another problem with the IT manegr's idea is that the dbms will
undoubtedly be dealing with real time data ? analysts using OLAP do
not like a changing data and thus will find it difficult to analyse
data.

on the other hand the IT consultant needs may have a challenge
implementing the data warehouse cost effectively


I can't see how the IT consultants idea is not the 'better' solution
considering he is using the DW with OLAP.

Again your feedback is greatly appreciated

Business Intelegance can be done only on consolidation the entire data
for a specific period in whihc i has to be seeked.
A database alone cannot do it. A Engine which can process the data for
various senario can only give you the desired business analytics.
Defining the various dimensions on which the data has to be analysed
is very cumbersom to handle. so you need a strong product with can
handle a multidimensonal structure processing.

dimenstion ( year X city X branch X customer)  something like this.

Hi. Thanks for the question.

Probably there are a large number of comparisons and discussions
between the various approaches.

Obviously, at the end of the day, it'll depend on your particular
requirements as to what's best. For example, if you have just a few
million records and only need reports on a daily basis, you'll want a
very different engine from one that works with 5,000 transactions per
minute and you need hourly reports on the trends across the database.

There are some good reports on Data Warehousing and OLAP available from the link:
* http://www.daniel-lemire.com/OLAP/index.html#2

and very large set of links (in Italian and English) is available from:
* http://www.olap.it/IndexEnglish.htm

Of course, if you are set on one particular database, eg Oracle, there
are documents on the appropriate database's web site, e.g.
* http://www.oracle.com/technology/tech/bi/index.html

I hope these links would help you in your research.
Please ask for a clarification if you have a question or if
you would need further information.

Regards,                              
R_a_t_t_y-ga                              
Google Answers Account Holder

No tip expected. We do it because we want to help.
If you want to thank us, either post a comment or just answer someone
else's question.

Consider, you have one PC

It has a large disk drive and a CD/DVD reader writer.

You have a load of data, should you keep it all on the PC, or shunt
the older stuff off onto a DVD in a slightly different format, mainly
different because the Indexes don't need to worry about allowing
immediate update.

A read only 'database' looks very different from a 'database' that can
be updated - you can have shed loads of 'short cut' indexes and linked
lists that would be very hard to maintain in a dynamic filing system.

The loss of read/write is not that serious, one can stick on a filter
that identifies any updates to ancient records that have not hit the
'read only' data section yet.

I agree with you, old data should be archived, multiple copies kept
(for security) and it should be optimized for searching and size.

It is also good for discipline, attempts to modify old data would show
up like a beacon.

The only 'catch' I can see in the question is that it might be
possible that the DW warehousing consultant is advocating ignoring
changes in dynamic data, eg: reporting on things as they were 24 hours
ago.

That would be madness, especially in a bank.

Your Alpha answer is that neither are right
- the answer is to implement both seamlessly

An illustration is:
'I want to run a report on our Eurobond positions exactly as they were
on 10th May 2006 at 14:05'

And now I want our latest (current) position.

Incidentally a 'Data Warehouse' should be renamed
- it should be called a 'Data Audit Repository'