Monday, May 13, 2013

SSIS Vs Store Procedure/T-SQL

One of the interesting debates/discussions which happen while discussing an SSIS implementation/Strategy is whether to use SSIS as a shell and do majority of the work the transformation/loading work done by SQL/Stored Procedures. The second option is to leverage the features in SSIS package and reduce the amount of TSQL/Stored procedures. In order to arrive at a optimum solution there are certain key points that need to discussed. The first would be the requirement from SSIS solution, in this there could be the following sub categories.
1. What would be the frequency of Data load into the target system.
2. What type of transformation is required.
3. What would be the requirement for the availability of data to the users.
4. What type of hardware is available at this point and how does it play into the requirement.

It is very important to understand the capabilities of the tools available while building the SSIS solution. Each piece of the ETL toolset would have a capability, positives/negatives, it is very important to understand them.
SQL/Stored Procedures are very good at performing SET based operations and CRUD operations. SSIS is very good at doing parallel processing, String based transformations, lot of functions for different type of operations based on data types, also it is graphical in nature. It is also very important to perform lot of testing while doing a SSIS implementation to get a good understanding of performance in the given server/hardware environment. With respect to transformations, get a good understanding of the complexity of the transformations required by the ETL solution. It would also be highly recommended to look at the metadata of the data sources and destinations and see if any design patterns can be arrived at. From a developer's view point it is important to understand the level of code maintenance involved.

In case you to get more insigh into SSIS best practices, please look into the discussion in Linkedin:
There are very good points listed in the link above by different SQL Experts, it is worth checking it out.

No comments:

Post a Comment