RCP job Vs Normal Job (Unix Server performance)

DS_FocusGroup · Post by **DS_FocusGroup** » Sun Sep 18, 2011 8:02 am

Hi,

Can anyone logically explain the difference on the load a Unix Machine has (in terms of CPU , memory and I/O) between a job that has RCP enabled and versus a job that has columns defined(Which one of them will consume more resources).If someone can give a logical analysis with some concrete analogies and examples then it will be great.

Thanks

chulett · Post by **chulett** » Sun Sep 18, 2011 10:33 am

I'd be surprised if there's any appreciable difference but that's just gut feeling based on nothing but what I had for breakfast. Be curious what the answer turns out to be...

qt_ky · Post by **qt_ky** » Sun Sep 18, 2011 2:57 pm

Are you asking about comparing, for example, an RCP disabled job with 20 columns defined vs. an RCP enabled job with the same 20 columns passed through?

chulett · Post by **chulett** » Sun Sep 18, 2011 4:32 pm

I assumed so... basically an "apples to apples" comparison.

DS_FocusGroup · Post by **DS_FocusGroup** » Sun Sep 18, 2011 6:25 pm

qt_ky wrote:Are you asking about comparing, for example, an RCP disabled job with 20 columns defined vs. an RCP enabled job with the same 20 columns passed through?

Yes for a start.

qt_ky · Post by **qt_ky** » Tue Sep 20, 2011 10:18 pm

I ran a test job just for kicks, so I can learn something too. Dataset w/ 10M rows -> Copy -> DB2 Connector (insert/replace). It's constrained to 2 nodes. One job has 13 columns defined and RCP disabled and the other job no columns defined and RCP enabled. I ran each job 3 times and on average each run took 2.5 minutes with the non-RCP job averaging 10 seconds faster. It's not a great test because of the short run times.

I can say the OSH from the RCP job is a lot easier on the eyes. I don't know internally what the difference is/how RCP works at run time, and will leave it to anyone else who wants to explain the internals or resource differences.

My guess is that the intention of the RCP feature is to save developer hours and not so much regarding resource or performance differences. I would want to see you run a test with a much larger data set to be sure, but until then I would have to agree that the performance difference is insignificant.