jump to navigation

SSIS for Extraction & Loading (EL) only September 2, 2013

Posted by msrviking in Business Intelligence, Data Integration, DBA Rant, Design, Heterogeneous, Integration, Integration Services.
Tags: , , , , , , , ,
trackback

There were series of posts earlier on the different ways to pull data from Oracle to SQL Server where I shared on how to port data either through Replication, CDC for Oracle, and few others like SSIS. I thought I will give a starter to a series of post(s) on how we went about picking up the option of implementing SSIS as a way to bring data from Oracle to SQL Server.

The major drivers (business and non-functional) for picking the solution were

  • The destination database on SQL Server should be replica of the source
  • No business rules or transformation need to be implemented when data is pulled
  • The performance of pull and loading should be optimal with no overheads or bottlenecks either on the source or destination

The destination system was mentioned as ODS, but it surely isn’t the same as Operational Data Store of DWH systems. Sadly the naming convention had been adopted by the shop where we had to implement the solution, and you might see me using the word ODS for sake of explaining the implementation, and I am sorry for putting the incorrect usage of the word. You would probably see little more of conventional or standards being skewed in the implementation, and of course a bit of my comments on how this could have been handled, better.

So the ODS was to hold data that would be used by downstream business analytics applications for the business users, with an intent of providing Self-Service BI. The ODS is to be of exact – in the form of schema, and data as in the source. The job was to pull data from the source without any performance overhead, and failures. Not to forget to mention that there needn’t be any transformation of data. At the end we use SSIS as only an Extraction & Loading tool instead of ETL – Extraction, Transformation and Loading.

This sounds simple, eh, but beware this wasn’t that simple because the design was to be kept simple to address all of the above factors. The SSIS package(s) had to handle these at a non-functional level,

  • The package should be configurable to pick and pull data from any set date
  • The package should be re-runnable from the point of failure
  • The package should have configurability to address performance needs – package execution time should be less, the source should not be overloaded, the destination should be full in its efficiency while data is loading.
  • The package should be configurable to pull data from source in preset chunks. The preset chunks could be based on a period basis – days /months /years, or number of records per period
  • The package should have the option to flag any other dependent packages to run or not run during initial and incremental loads
  • The package should have defensive design to handle bunch of different type of errors
  • The package should have error logging and handling at a package level, and at record levels

In Toto this solution was more to how do we design for the non-functional requirements and implement, leaving the functional side of the data. This type of implementation is half-blind and I will talk more on those cons we had to face, when I get in to details of each implementation.

A post after few weeks of gap, and one with learnings and summaries on how it could have been better.

Advertisements

Comments»

No comments yet — be the first.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: