DSXchange

While running the job, use your favorite monitoring tool and see if you are peaking on disk I/O, CPU or Memory.

it is at "/opt/IBM/IIS/InformationServer/Server/DSEngine/lib/libgciudt6.so
" on my AIX; so should be under "$DSHOME/lib"

That approach sounds right. What character value is this being converted to? You can use SEQ(In.Character) in a transform stage, or output to a text file and use a UNIX utility such as "od -x" to see what the numeric value is. It would also be good to know what character is used for that o...

The environment value you are looking for is $APT_MAX_DELIMITED_READ_SIZE and its use is documented in the advanced parallel job developer guide.

Here in Germany, "PLZ" is the zip code, or "Postleitzahl".

With that type of volume and unsorted data, you are going to use disk I/O. Is your scratch disk local (good) or on a SAN (bad)? Are you using unbounded VarChar lengths (good) or explicitly set lengths (not so good)?

I can't check the different screens, but I think you are on the right path in looking at the Suite Admin settings - unless access is given correctly that button will not detect you as administrator and remani greyed out. Could it be the roles are set up, but not assigned?

The relevant options are "sort Utility" and "Stable Sort" and "Restrict Memory Usage". You should leave the buffering mode alone (initially). How many nodes are you using in your Dataset? Can you use more nodes? Is the data that is written to the Datasets already sorted...

Sorting partially sorted data tends to be faster than sorting random data. You can insert a sort stage with "Don't sort, previously sorted" for the already sorted columns to let DS know that it doesn't need to sort data.

Those factors don't affect sort performance. What SORT stages attributes have you set or changed?

Is this a sporadic error occurring at different positions? If so, this might be the same error that we have on this and I have a patch from IBM (that I still need to test). It seems that the generation works correctly, but that it produces that spurious error message.

And how about "oudata1" or is it "outdata1"?

And how are these variables declared?

What are source lines 40-50, could you post those?