Impacts of changing Buffering Policy

pavankvk · Post by **pavankvk** » Mon Nov 09, 2015 11:11 am

Hi,

We are upgrading from DS8.1(Linux) to 11.3 (Windows). We have quite a few jobs that use a design similar to fork join. The jobs are getting into a deadlock situation and stuck with no progress. These jobs are running fine in 8.1(linux)

But when we change the buffering policy from automatic to force, they are running smooth. What is the impact of forcing buffering, with out the tool deciding it. we have other jobs also, that ran fine, so would forcing buffering to every job/link cause performance issues?

Has anyone come across this scenario? Planning to open a ticket with IBM,but wanted to get an expert opinion here.

testuser33 · Post by **testuser33** » Tue Sep 06, 2016 7:48 pm

Hi,

We are having similar issues on DataStage 9.1. Could you please help with what was recommended and on the impacts due to this.

ray.wurlod · Post by **ray.wurlod** » Tue Sep 06, 2016 9:22 pm

There may be no impact, if the automatic buffering policy was already inserting buffers.

Worst case scenario - if you set APT_BUFFERING_POLICY rather than per-stage buffering policy - is that you will need sufficient memory to accommodate buffers for every link in the job. By default there are two buffers each 3MB per link.