Performance optimized Techniques /concepts in Teradata

General covers Articles, Reference documentation, FAQs, Downloads and Blogs that do not belong to a specific subject area. General-purpose Articles about everything and anything

Performance optimized Techniques /concepts in Teradata

Hello Everyone,


Can anyone explain me the performance optimized Techniques in Teradata?


Thanks in Advance!


Senior Apprentice

Re: Performance optimized Techniques /concepts in Teradata


No disrespect, but: "how long have you got?" There are a variety of techniques which provide benefits in differing circumstances.


To give you some starting pointers:

  1. Remember that Teradata's architecture is based on parallel processing and this underpins practically all thoughts about performance.
  2. Always know what sort of volume of data you're dealing with whether that is loading, processing on the TD system or extracting.
  3. Except for small tables, choose the Primary Index of each table to provide good data distribution. I aim for parallel efficiency (a measure of even distribution) to be 95-95% or higher.
  4. Avoid row at a time processing although this normally happens with cursors in stored proc's.
  5. Aim to load the data into TD at the earliest possible point in your processing and then keep the data on the system at all times. Normally referred to as ELT rather than ETL.
  6. Avoid extracting the results of some processing to an application server, doing some transformation and then loading the transformed data back.
  7. You will typically need statistics on every table even if it is only SUMMARY stats.


I think those are some good generic starting points but I doubt if any of them are 'mandatory'. There are probably situations where you can safely ignore one or more of those.


Anything specific that you were interested in?




Ward Analytics Ltd - information in motion