DSXchange

Define large numbers; i.e. some languages have "one, two, three and many", so 4 would be a large number

that having been said, look up EXACTNUMERIC if your problem occurs with number of greater than 15 digits.

Do you have a tool such as MQJexplorer or similar available to you? I suspect that only 99 characters are being put into the message, as the MQ Stage can certainly read quite a bit more than that.
Try writing messages to the MQ Series queue using a Datastage job, that might work better for you.

Go to your output sequential file stage and column view, double-click on the column number for the nullable field and see if the column meta data actually does show that you are handling nulls.

Unless your Database is partitioned, going for more nodes in your configuration file is not going to make a positive difference. So far the only size you've mentioned is the 35Gb of scratch. Since we have no idea of the sizes of the two data sources it is impossible to make any suggestions or progno...

What was wrong with the sample code that I posted, that should work.

Performance is more a matter of how fast your network connectivity between the two machines is, and how loaded your DS server is.

In Version 7 you can still use the contents of the DS.AUDIT hashed file to get that information, as of version 8 that information is hidden in the bowels of XMETA.

I think that the exactnumeric change to 19 ought to do it.

Once your data for left and right parts of join is sorted on the join keys then your job won't need excessive amounts of scratch space. It will need scratch space for sorting. You didn't mention how many Mb/Gb your two data tables are.

I'm not sure if this applies, but what is your uvconfig file value "EXACTNUMERIC"?

In the sequential file stage, just enter "bzip2 -c" to compress using bzip2, other commands have slightly differing sysntax to take their input from stdind.

I wouldn't use the compress stage for this. Just use a sequential output stage and add the command "gzip -i" to it . I'm not at a DataStage computer now, but when I do I'll send the exact syntax I use.

I've been working with MQ and wave processing at V8.0.x for over a year and wave processing works.

I'm not clear from your description on what aspect of wave processing isn't working?

One job parameter can only have one value. What do you mean by a set of values?

Without a version control system there isn't a way to do it, unfortunately. I haven't looked into what has come out at V8.1, but there might be something there; V8.0x has nothing.