server and parallel migration

tostay2003 · Post by **tostay2003** » Thu Mar 09, 2006 8:55 am

Hi,

I am about to join a company which has newly acquired 7x version. They were currently working on server edition. My tasks will be to work on both parallel. I presume it could be migration as well. Dont mind my basic question. But could u pls shed light at elementary level how the migration process will occur (what about the hash stage, as it is not used in parallel). What could be role of an Admin and ETL developer in this regard.

ArndW · Post by **ArndW** » Thu Mar 09, 2006 9:38 am

You can't migrate hashed files to PX; you need to redesign the jobs. The same re-design applies to almost all Server jobs in order to take advantage of the parallelism possible in PX.

The migration should be approached assuming the products are quite different, i.e. you will need to do some project planning. Most of the time the re-development time won't be worth migrating. It can cost just as many people-hours to redesign an existing Server job into a PX job as it did to originally design it. Plus you need to go through the whole testing system as well with the new job.

tostay2003 · Post by **tostay2003** » Thu Mar 09, 2006 10:32 am

So u mean the existing server jobs should be left alone. And the new jobs taken up should be done in parallel.

One small doubt, is it that we just need to open the .dsx files of server jobs in 7x version and compile them (If we dont want to redesign). What about the environmental variable etc. Is it as simple as it sounds, just opening the .dsx files in 7x version or are there anymore precautions for it?

ArndW · Post by **ArndW** » Thu Mar 09, 2006 10:38 am

The PX install will be an add-on if it is at the same version, so you won't need to do anything with PX jobs.

If you spent $1000 on designing a job that works acceptably, it will take a powerful business case to justify another $1000 in making it into a PX job which does nothing else except to run in half or a third of the time.

tostay2003 · Post by **tostay2003** » Thu Mar 09, 2006 11:17 am

i wasnt able to see ur message with that premium content.

ray.wurlod · Post by **ray.wurlod** » Thu Mar 09, 2006 11:42 am

Server jobs will be around for the foreseeable future. This was reinforced in (unofficial) discussions with IBM technical person this week. The first question you must ask is whether there's any value in migrating, given that fact. If your server job performs well, and meets your time window, it's not broken, so don't fix it. If it doesn't meet your time window, consider converting it to multi-instance and running two or three instances in parallel.

tostay2003 · Post by **tostay2003** » Fri Mar 10, 2006 7:56 am

thank you

mystuff · Post by **mystuff** » Mon Apr 16, 2007 8:16 am

I heard that we can use server shared containers to call server jobs.

If I have a sequence, where Sequence2 is calling server job & rest calling parallel jobs.

Sequence1 - Sequence2 - Sequence3

Then will Sequence2 become bottleneck in performance?? Or will it run with parallel computing ability?

ray.wurlod · Post by **ray.wurlod** » Mon Apr 16, 2007 3:21 pm

Etiquette Note
Do not hijack threads. This question is not sufficiently closely related to the original subject. Please begin a new thread.