Overview
Architecture
Application Development using TD
Performance Considerations while using Teradata
Q&A
OverviewOverview:
The Teradata RDBMS is a complete relational database management system.
Features:
-> Parallel Architecture
-> Compliant with ANSI SQL.
You can access data using standard SQL.
TD provides standard ODBC driver, which means you can access teradata with any ODBC
front-end applications. Eg: TD SQL Assistant
TD is accessible from various platforms
* Mainframes
* Unix
* Windows
and third party tools and applications like:
* SAS
* BI Tools: AbInitio/Datastage/Informatica/Sunopsis/Cognos/BO
Teradata native tools:
TD SQL Assistant -- For Developers, Business Users
BTEQ, Multiload, Fastload, TPT, Teradata Administrator --> For Developers
Diagram: Typical Teradata Query Processing:
Architecture:
Insert a picture of Teradata Architecture here.
Application Development using TD
Teradata is a standard RDBMS. Most frequently used TD objects for development are:
Databases
Tables
Views
Macros
Note:
It also supports Stored Procedures. However there usage is discouraged due to their nature of
sequential processing and impacting performance.
Understand Primary Index in Teradata. PI is access path as well as distribution mechanism for
parallel processing.
Teradata supports ANSI Standard SQL along with Teradata Extensions
Typical Teradata application involves :
Load: Uses Teradata Fastload (or TPT Load operator) and load data into a Teradata staging
table
Transformation: This can be done using native Teradata SQL using views and BTEQ job scripts,
using ETL tools like Informatica/Datastage and load tools.
Perofrmance Considerations:
1) Use Primary index in join conditions. If PI is not used then this needs to be holistically reviewed
at data-model level.
Eg: In PARTY subject area, expect that all PARTY related tables should have the same PI:
Party_ID
2) Ensure that statistics are being refreshed on periodic basis.
3) Explain the sql and see where you can improve performance.
4) Use Teradata Load operators while loading voluminous data into Teradata.
Eg: When importing a source files into staging area, use Teradata fastload or use ‘Bulkload’
operator in your ETL Tool (eg: Datastage)
5) Insert-Select:
Insert –select is optimized and fastest way of copying data from one table to another table. In
a typical data warehouse implementation we can see data being transferred from ETL tables to
EDW Target tables using this method.
6) Split the SQL: If your sql is too big and complex then split into manageable modules and join
them together
