Archives of the TeradataForum
Message Posted: Wed, 10 May 2006 @ 15:59:13 GMT
| Subj: | | Re: Help me in finding answers for few questions |
| |
| From: | | John Graas |
A lot of information:
| | 1) How partitioning a big table can help improving the performance in Teradata? | |
Primary Index Partitioning (PPI) helps to eliminate data blocks (contained in eliminated partitions) from query data scans, lessening I/O and
CPU requirements.
May I suggest the white paper at
www.teradata.com
for this question and the next?:
www.teradata.com/t/page/104071/index.html
| | 2) What is the best practice for managing a huge table? Should we split one big table into multiple tables (on the basis of one
attribute, for e.g. Year) & use the latest table to answer the query. | |
| | We can hold all of the history data in one big table & use one table to hold the latest data. For example to store the data for last 5
years in one table & for current year we can use one table. | |
My preference is one table. If the PPI matches well to the query access then this is pretty much a given. For example, if the PPI is date
based and the majority of queries qualify date, then one table is the first option.
| | 3) What is the best practice in using Staging server? Should we use multiple small staging servers for Extraction &
Transformation or should we use one big server to handle everything? Which one can be easily managed? | |
Most of the accounts I have dealt with use one staging server to support an ETL (Extract, Transfrom and Load) tool, and performance has been
fine. If you have a huge transaction volume, look for references from your ETL tool vendor for designs at that size just to make sure. However,
if you are considering an ELT (Extract, Load and Transform) design then the Teradata platform should be able to handle the staging duties.
jdg
www.jgraas.com
| 
