DSXchange

Under what user ID do your jobs run?

Not necessarily. Read through the offered solutions again.

The most obvious setting is the maximum number of instances of a multi-instance ISD job.

Lookup File Set is always totally overwritten. Examine the score; you will see a composite operator that deletes the LFS then populates it (including building the index). The reason for that is that the index is a sorted hash table (sorted on the Key) that includes pointers to the physical location ...

Check that you have exactly the same file name on both input links to the Hashed File stage - not only identically spelled, but also identically cased. Incidentally, your link row count is 1.7 million, not 176K.

There are no three dot characters in European languages, though there is a couple in Arabic: thaa (ث) and shiim (ش). Arabic does not have a "v", so sometimes three dots are placed over faa (ڤ) to represent this sound. Still other "three dot letters" are used in North African and ...

I would imagine so, by having some dynamic form of resolution from name to IP address. What name resolution technology do you currently use?

This (qt_ky's suggestion) is the exact purpose of the Column Export stage.

You might like to investigate the Server to Parallel training available from the DSXchange Learning Center (a commercial site).

Mike's memory is good. Handle the nulls in your stage variable expressions such that the final value of the stage variable is not null (for example using the NullToValue() function or one of its brethren).

Might not that confuse your users? For example, today they find a term definition (connected to A) but tomorrow it's not findable (because they are connected to B). How do you propose to synchronise your catalogs?

I wonder if you could use an External Target stage using the cat command and runtime column propagation, and a collection algorithm that exhausts each input before moving on to the next?

Check that your C++ compiler has a current licence.

Try using explicit, rather than implicit, data type transformations. And be careful about null handling.

What's different between the development and production environments?

Hint: "nothing" is not the correct answer.

Tell us more about the job design and about how the job is invoked from the sequence.