DSXchange

Hi,

my job starts with a flat file and enriches data by 20-30 lookups.

For performance reasons I have selected the data into hashfiles so the lookup source is alsways a hashfile.
Because I have so many lookups and hashfiles the configured memory is not big enough to hold all data in RAM.
The option pre-load to memory is check in every hash file but the log of cause shows that it can not be cashed for all of the files.

Any ideas how I could increase the throughput for that job?
Currenty it is processing 30-40 rows /s.

I tried to split the job into smaller ones with 10-15 lookups with no success.
I tried to use multiple processes for that job - with even worse performance.

Which memory configuration sould I increase for the best performance increase?

Any help and ideas are appreciated.


kind regards

Michael

Try using four or five Transformer stages, with four or five lookups each, and enable inter-process row buffering. That way you will have four or five separate processes each with its own memory allocation.

Could you post both the error message you get and your system virtual memory setting?

ArndW wrote:Could you post both the error message you get and your system virtual memory setting?

Hi Arnd

I do not get any ewrror - it is just too slow

The memory configs are default so far

kind reagards
Michael

My guess would be that your hashed files are ineffeciently sized or too big. You need to also monitor your CPUs and see what the job is doing. If other things are running you may be contributing your poor performance to the wrong reason.