Archives of the TeradataForum

Message Posted: Wed, 26 Mar 2003 @ 14:55:32 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

Narayan,

What specifically are you storing on these systems.

1) are the databases identical, yet are populated from different systems (same structure - different data)?

2) are the databases replicated? exact duplicates, data and structure?

3) are the physical databases designs/structures different?

3a) are the databases both "data warehouses"?

3b) are you using one large "data warehouse" and several smaller "data marts"?

3c) are the databases mixing a data warehouse and OLTP systems?

4) are all or both databases Teradata? or are you "mixing" RDBMS platforms?

Although, as David Wellman pointed out, NCR may or may not support the capabilities that you desire. Before you are told that this is not possible, it is important to understand what you are trying to do. How the data stored and what your data access requirements are.

There are ways to "virtually" integrate multiple RDBMS'. We, in fact, delivered this capability to NCR as part of TeraCube.

We have several customers that have "virtually" integrated multiple RDBMS'. One of our partners is building a multi-petabytes, GRID based "virtual" database. This may be beyond your design requirements, my point here is that there are some very interesting efforts underway that may address your concerns (www.ggf.org and www.globus.org have additional information).

As you may expect, very large joins between the systems are not very efficient, but they are supported. Otherwise, the systems "appear" as a single physical source and therefore are accessible with "off-the-shelf" query and OLAP tools, and scripting languages.

God Luck with your investigation.

John Grace
Geppetto's Workshop L.L.C.

	<Prev	Next>		<<First	<Prev	Next>	Last>>