>From a technical perspective this is not difficult.  From a  political
perspective it might be a can of worms.


Before wasting time on technical requirements and capabilities you will
need to have the support from all the contributing organizations and the
support of the relevant IT staffs.  Depending upon how autonomous each
organization is may affect your ability to "pull this off".

This type of project will require that data be "copied" from the source
system to a target system.  In doing so you will require that "scripts"
be run against each of the contributing databases on some regular basis,
the people who manage these databases may have some issues with this
(calm their fears now).  If it is a single organization you are dealing
with you will most likely have better results.

Inventory ALL the data sources, obtaining detailed information about the
systems, software and data structures in use and any pending changes
that may be made.

For the "technical steps":
1) Develop set of specification that describes the FUNCTIONAL
requirements of the project. 
NOTE: You will need to decide what type of information you want retained
in its original detail and which can be aggregated.  Be aware that once
aggregated you can only get back to the detail be restoring the detail
data, assuming it is still available.  Keeping the detail, at first
glance, may seem the way to go.  Be analytical about what really need.
In general, more detail means more storage and CPU power for processing,
and more time  for loading, and backups - more overhead.

2) Take the functional requirements and develop a set of technical
specifications.  These should contain specifics from storage structures
to analytical requirements.  You will need a "first level" data flow
diagram and entity relationship diagrams.  (what I mean about first
level is that the specification should be system independent, not
relying on a specific database feature, unless the database has already
been decided upon).  This should also contain a set of procedural
requirements in your case (the when and how data will be ported to the
warehouse)

3) Evaluate your technical specs against the available DBMS system you
will consider using.  
Note: this may be your first political motivated decision.  There may
already be a DBMS requirement that you will need to conform to.

Assuming you will use MapInfo as your mapping client then you have the
following DBMS choices:
Oracle (using Oracle  Spatial)
SpatialWare on : Informix or SQL Server
(I will not interject my biases here)

Each will work well within the MapInfo family from MapInfo as a fat
client, to MapXtreme for web based applications. Each having direct
access to the warehouse. 

You may want to consider the  geocoding capabilities of each
environment.

4) Depending upon you project management / development methodology this
will be the time to start prototyping.  Design a reasonable testbed and
measurements for evaluation.

There are a bunch of other steps I can't elaborate on here...

Some final thoughts:
Expect to cycle through the design process at least three times before
arriving at something that seems to accomplish what you want.

Data Warehouses DON'T occur over night... double your best estimate.

There are TWO "people" you want to find right away - There will be at
least one supporter at a high level (if not, can the project).  There
will be at least one person against the project - know who they are and
how to deal with them.

Your first implementation will not meet everyone's needs.  Let people
know what is intended to be accomplished and what is anticipated in the
future.  In other words, publish a project plan and keep it updated.  (I
have always let people know about failures as well as successes).

That all I have time for now
Guy Groves
GRG Consulting


-----Original Message-----
From: Norman Mabunda [mailto:[EMAIL PROTECTED] 
Sent: Monday, February 02, 2004 06:19 AM
To: [EMAIL PROTECTED]
Subject: MI-L Data Warehousing

Hi All

This will be answered probably by the database experts, however it's not
a difficult one. I have been assigned in my work to research on ways to
develop all the national health datasets and create a central inventory
or repository for this datasets for all 9 provinces in the country. Data
Warehouse is the concept being used here. Who can best define and
explain what he thinks of:

- Data Warehousing
- Data Warehousing process
- Data Integration
- Data Mining
 - ETL
- OLAP

What would be the most effective way of developing data warehouse.

Regards,

Norman


---------------------------------------------------------------------
List hosting provided by Directions Magazine | www.directionsmag.com |
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
Message number: 10231

Reply via email to