I have a question about correct design/performace for the data source in SSIS.
I am moving data from some oracle tables to sqlserver database on a regular
basis.
Some tables only have a few thousand records but 1 has 2 million another has
4 million records. Based on my current understanding I have placed all of
these data moves into 1 data source so that they can be executed in parallel.
I am presuming that this will give me best performance. After all of data
is copied I then execute some execute sql task on individual tables. I do
not care what order the data is copied from remote oracle database.

So my question is this the best way to design this part of application? Did
I make any bad assumptions? Is their any reason that I should put each copy
data into its own data source?

Hello brian_harris,

If you do not care about the order of the data being loaded then I would
be tempted on a multi proc box to divide my source data into ranges. Let's
take time for instance. Divide the source data up into equalish chunks.
How many chunks? I generally start with CPU Count - 1.

Now in either a single Data Flow Task with CPU Count - 1 Pipelines or CPU
Count - 1 Data flow tasks with 1 pipeline in each you move that data in parallel.

Donald farmer gave a presentation the pther night on Perf. You should be
able to download it over at MSDN Webcasts


Allan