DSXchange

No.

Whatever you do, you are going to require two passes through the data to do anything with ranks.

Do you really need a Link Collector? In theory your job should go faster with parallel independent streams of inserts.

Loading the required rows and columns (and only those) from the DB2 table into a hashed file and using that for the lookups will be fastest.

What character map(s) have you set? TH-THAI is a locale setting, and has little if any bearing on the display of characters. You need to specify the correct character map on any passive stage (one that interacts with persistent data).

The only "priority" is that they are executed in the order in which they appear in the grid; reordering the grid rows is a simple matter of dragging the leftmost column (stage variable name) up or down as required.

IMPORT the table definition from Oracle. What comes in?

One of the strange things about functions in DataStage BASIC is that they can, indeed, be called as subroutines; the first argument position is reserved for the return value. Therefore the following are equivalent. Call DSX.UTILITYHASHLOOKUP(Result,"HA_REC_TYPE","2000",2)...

What do you want to get from DS_JOBOBJECTS pertaining to parameters?

Change CUST_ID to Integer in your job. Oracle uses NUMBER, which imports as Numeric, but SQLColAttributes() understands ODBC conventions rather than Oracle conventions. This can lead to minor inaccuracies when importing metadata from Oracle tables. You must make the metadata match.

You have clearly missed at least one vital aspect of those permissions.

It is not offensive. In my own terse way I was trying to make a point that no-one should be asked to adminsiter DataStage without vendor or good third-party training. You didn't become a DBA without training, did you? DataStage can be more complex than administering a database. DSXchange is here to ...

Only until you exhaust available memory for the hash table, then it must spill to disk, which is quite a slowdown.

An internal format does exist for date, time and timestamp in parallel jobs; they are all binary and the various StringTo... functions (Transformer stage) or ..._from_string functions (Modify stage) effect the transition if what you have is a string.

Or run the Transformer stage in Sequential execution mode.

Slightly off topic, the decision about Transformer versus other stage types is also a function of version. Somewhere in version 7.1 Transformer stage performance was enhanced, and can now be preferred to Filter or Switch (for example) which are effectively interpreted rather than compiled. However, ...