Summary: SSIS Data Flow Task in Katmai makes better use of multiprocessor machines . . .

Out with the Old …

The SSIS Performance Tuning Article explained the performance tradeoff between Worker Threads and Memory Buffers in the context of Execution Trees. Essentially, the designer (you) would test several design scenarios where each design was created to take advantage of more worker threads (parallelism) or fewer memory buffers. You would then determine which techniques worked best for your application and implement that design.

Unless you get very creative, you may not take advantage of the available CPUs in a high-end multiprocessor machine.

In with the new …

In Michael Entin’s Weblog, Michael reports that the new Data Flow Task in Katmai will use parallelism within a single Execution Tree. Michael is a Technical Lead on the SQL Server Integration Services Team.

What this means to you…

In Katmai, you can skip the techniques mentioned in the SSIS Performance Tuning Article mentioned above and get more parallelism in your Data Flow tasks right out of the box!

Jim Leisring, Scalability Experts