Migration from oracle to teradata

[seethamsetty]

Hi!

can any one throw their experiences, in migrating the existing Orale based DWHS to teradata

ETL used Datastage,

Best approaches , or will be good if some one can help on documentation

Thanks

[ArndW]

Such a migration project will, of necessity, involve a lot of thought about reorganizing the tables in order to take advantage of differences in architecture. Since those changes will also affect the ETL layer the real answer is that a migration from Oracle to Teradata will never be a simple one; and by extension the project plan for a migration will be complex enough that the tasks cannot be listed here - particularly as there are so many elements involved.

If you want some answers to specific topics you would be best served by narrowing down your question.

[chulett]

Such a migration project will, of necessity, involve a lot of thought about reorganizing the tables in order to take advantage of differences in architecture.

Bah... one part of the organization here is in the process of same. First step? The migration of the Oracle database to TeraData unchanged as much as possible.

[ArndW]

I guess that keeps the {initial} migration costs down

So all you need to do is go into the Designer for each job, drop Teradata stages onto the canvas and move the link over from the original Oracle stages. Fill in a couple of parameters and Bob's your uncle. Why hire specialists for that, just get some high school kids to do it for you at minimum wage plus a couple of percent and you are done.

At this rate we are going to be out of migration consulting jobs for DataStage. On the other hand, it will keep us employed when the "new improved" ETL ends up being no faster than the original it was built to replace and they start looking for performance or load improvements

[kduke]

I have always liked Oracle even preferred it. Teradata is smokin fast. The different stages are not like any other database. Somethings can be slowed down by how the DBA sets up the database. There are some really good training docs by Teradata to explain the difference. Read them first.

[vmcburney]

Your two biggest challenges:
- the different way that Teradata handles keys and indexes. It doesn't need a unique primary index - in fact it can perform better without it. It does like natural keys instead of surrogates. You really need a Teradata specialist to get this right. An Oracle data modeler will almost certainly botch a Teradata physical data model.
- the different ways Teradata runs inserts and updates. Not as simple as Oracle or DB2.

You may have to add or remove database insert / update stages and test and evaluate the different load options on low, medium and high values across simultaneous parallel jobs until you are on top of the strengths and drawbacks of each method.

[hamzaqk]

Teradata does have indixes. and by default if no index is define it does create a primary index on the first column of the database. it lets u define 4 kind of indixes. Primary index, Unique Primary index, Secondary index and Unique Secondary index. So yes it does allow UPI and it does help in improving the performance as all the internal data distribution in TD is based on ur column selection and how unique they are

Teradata Certified Master V2R5

[hamzaqk]

Teradata does have indixes. and by default if no index is define it does create a primary index on the first column of the database. it lets u define 4 kind of indixes. Primary index, Unique Primary index, Secondary index and Unique Secondary index. So yes it does allow UPI and it does help in improving the performance as all the internal data distribution in TD is based on ur column selection and how unique they are

Teradata Certified Master V2R5