DSXchange

Generate the data file locally on your DataStage machine. Transfer it to your database server machine, using any appropriate command (depending on share protocols you may be able to use cp , rcp , ftp , and so on). Presumably you have the bulk loader control file already on that machine, so you can ...

While it is possible to tune hashed files, I would like to see your proof that it is the hashed files that are to blame. Hashed files are a very efficient mechanism for performing lookups - probably the most efficient. So, please post your proof, then we'll see about your assertion. I suspect it's s...

An attempt to write beyond the 2GB boundary of a 32-bit hashed file will result in its being corrupted. You will get backward link (BLINK) errors when trying to access the illegal address. Presumably an attempt to write beyond the 19 million TB boundary of a 64-bit hashed file will result in the sam...

If you think about it, you will see that the answer is derivable. There are 86400 seconds in a day. 86400 + 7200 = 93600. What you need to incorporate is a test that the end time is earlier in its day than the start time is in its day, and subtract 86400 in this case.

Which test did you run first? What steps did you take to control for the effects of caching - that is, the required rows being cached somewhere in the database server?

There is no upper limit on T30FILE explicitly. All of the structures sized within uvconfig must fit into a shared memory segment (sized by shmmax kernel parameter). T30FILE sets the maximum number of slots (rows) in a table of open dynamic hashed file information in memory. This table can be viewed ...

Please post all settings you have used in your UniData stage. When you refer to "repository", I trust you don't mean DataStage repository (which is not based on UniData). Can you create a completely new job that consists only of a UniData stage and a Sequential File stage? This will allow you to nar...

That message has nothing whatsoever to do with indexes. It relates to the fact that the design-time information is missing. This can come about, for example, by exporting only the job executable - not the job design. It can also result from partial deletion of the job design. You need to establish w...

Surely what you're already doing ("every 1st") covers what you've asked?

When Ascential acquired Torrent Systems they renamed the parallel execution technology from Orchestrate to Parallel Extender. That name has now been dropped in favour of Enterprise Edition.

The easiest ways are to have Job A write the values to a file and have the job controller (sequence) read that file before invoking Job B, or to have Job A load the values as a delimited string into its user status area and have the job controller access and parse that value before invoking job B.

SMP is "share everything" - all CPUs in the one machine share access to the same memory resources, the same disk resources, the same drivers, and so on. In DataStage EE, inter-process communication takes place via shared memory. MPP is "share nothing" - a number of independent machines each having i...

Well done! I'm sure we would have gotten there, systematically, eventually. Sounds like a good opportunity to use a job parameter!

You really should leave the indexes alone unless you have reason to believe they need rebuilding.

On an offnote, I love using "CustomOPS" and "BuildOPS", and watch everyone else sputter and stumble on the difference between the two. Co-worker: "So you can compile the CustomOPS in DataStage?" Me: "No, but you can generate BuildOPS." Co-worker: "But you have the source code for BuildOPS outside D...