SELECT * From SSIS.DataFlow

If you’ve been looking through the documentation for the Power BI preview, you might have noticed a section on Publishing SSIS Packages as OData Feeds. This functionality lets you create a T-SQL View over an SSIS data flow using a new SSIS add-on called the Data Feed Publishing Components. This add-on works with SQL Server 2012, and is a free download from the Microsoft download center. While the components are useful for a number of Power BI scenarios, the components don’t require a Power BI subscription – all you need is a SQL Server/SSIS 2012 installation.

The Data Feed Publishing Components have three main components:

  1. Data Streaming Destination – A new Data Flow Destination that creates an “endpoint”, letting you stream data back to the caller (similar in concept to the DataReader Destination)
  2. OLE DB Provider for SSIS – A special OLE DB provider that allows SQL to treat an SSIS package as a Linked Server
  3. Data Feed Publishing Wizard – A wizard that deploys a project/package containing a Data Streaming Destination and creates a T-SQL View (and linked server) that kicks off the SSIS package when accessed

Publish SSIS Packages - Conceptual Diagram

If this sounds interesting to you, be sure to check out the step-by-step guide in the Power BI documentation. Note – if you’re currently not in the Power BI preview, you can stop at step #3.

Scenarios

I expect I’ll be blogging more about this in the coming months (as well as talking about it at PASS), but I wanted to briefly mention some of the main scenarios we’ve been working with customers on.

Invoking an SSIS Package from a Report

You’d do this in the case where a simple query isn’t enough – there are work flow steps (i.e. FTP files from remote server), you’re merging/transforming disparate data sources, require .NET scripting logic, or your data source requires a custom connector. Internally we’ve been referring to this scenario as “Complex Data Feeds”.

While it is possible to configure Reporting Services to read from an SSIS package, the approach has some limitations as to the account the package will execute as (and is actually removed from the default configuration file in SQL 2012). The Data Feed components provide a similar approach, but also let you leverage the logging and configuration provided by the SSIS Catalog.

On-Demand Execution of an SSIS Package

SELECT’ing from the View created by the Publishing Wizard dynamically invokes the SSIS package with a data flow, and streams back the results. While the majority of SSIS packages would run on a schedule, or write data to a fixed destination, there are cases where dynamic invocation and streaming results are preferred.

One customer we worked with had 500+ “Data Feeds” – data sets that were more than just a simple queries. This data sets were typically small and used for ad hoc reporting purposes. These feeds weren’t accessed regularly – some would not be used for months, and then be used heavily for a day or two (perhaps at the end of a quarter). Unfortunately, the access patterns weren’t predictable. Because the data had to be there when it was needed, the customer ended up with a very large infrastructure keeping every feed up to date. What they needed was something that could be run on demand (perhaps with built in caching), so the data could be refreshed only when it was needed.

Another customer was looking for a way to do dynamic auditing in their environments using SSIS. They had a set of packages with custom auditing logic that they’d deploy and run in various environments, getting a real-time snapshot of their systems.

Alternative to Linked Servers

Want to use a linked server, but don’t have an OLE DB provider for your data source? Want to enforce custom logic, or do dynamic transformations on the results of the data? No problem – use an SSIS package!

Limitations

Just like Power BI, the current release of these components is in preview, and might not have all of the functionality you’re looking for (just yet). One thing to note is that the SSIS OLE DB provider currently does not support Distributed Query optimizations. Therefore, it currently doesn’t provide statistics or push down filters like other OLE DB providers used for Linked Servers. This functionality is best suited for one time executions of an SSIS package – if you find it’s something you’re accessing over and over, then you should probably be running your package on a schedule.

Links

For more information, see one of the following:

One Response to “SELECT * From SSIS.DataFlow”

  1. […] SSIS packages as data sources.  In the meantime, as a workaround you can use linked servers or SSIS Complex Data Feeds to expose other sources).  Once it is installed, it establishes a communication channel with your […]

Leave a Reply

You must be logged in to post a comment.