Archives of the TeradataForum

Message Posted: Mon, 26 Sep 2005 @ 18:46:58 GMT

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

Rishi,

The FS-LDM makes extensive use of system generated keys (SGK), so while the UPI typically appears to be on the PK, this is simply because the PK is the SGK (anyone wishing to begin a long and involved philosophical discussion of generated vs. natural keys, please try to resist). There are a number of very good reasons for using SGK's, but for my money the most compelling is the space consideration involved in migrating the natural key data to each row of every dependent table.

It's important to remember that in Teradata, the main purpose of the primary index on a table is distribution, then all other purposes come after that. The best PI is one that produces a relatively even distribution across the AMPs, which helps to ensure efficient query processing in parallel.

My experience with the FS-LDM has been to:

- make use of SGK's, especially for the main entities of party, account, event and product

- maintain the natural keys as data attributes on the main entity

- keep the model as normal as possible

- make use of aggregate tables for account balances

- pay close attention to the relationship tables (i.e. party to account, account to event)

- employ secondary indices to improve query response time

- use Teradata PS to get you over the initial modelling hump

Hope this helps,

JL

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