Page 1 of 1
Parallel vs Server
Posted: Wed Apr 27, 2005 9:42 am
by Mark_E
hi, somebody posted a while back the difference between a parallel job and a server job. can anyone remember the link. or explain the differences between the two and benefits.
thanks in advance
Posted: Wed Apr 27, 2005 10:30 am
by ArndW
Mark_E - this question is even broader than your other current thread. Px and Server have a common front-end but very different technologies under the covers.
A Server job will execute within the DS environment (the erstwhile UniVerse engine) and use it's advanced string manipulation and conversion features to great advantage. Just about any type of transformation is possible with minimum effort and a steep learning curve (for the BASIC language that underlies the whole system).
A Px job does not use the Server engine and thus does not profit from those items mentioned above. It uses C++ generated code and routines. What it does extremely well is to implicitly run whatever it can in parallel. It is very, very fast and scaleable on SMP MPP systems which is something that the Server jobs can only do with a lot of explicit coding; and some of the functionality in Px is not available in Server at all (i.e. you can design a generic Px job that will accept input files of just about any column definition)
Posted: Wed Apr 27, 2005 7:39 pm
by trokosz
Well big issues are....
1. Parallel Jobs require the C++ compiler for your OS installed and all its load libraries....The Transformer Stage needs this....
2. Parallel Jobs do not know about Hash Files.....
3. Parallel Jobs need scratch and temp space to manage its partioning and nodes....
4. Parallel Jobs provide several more Stages not available in Server like Debug, Row Generator, Copy, and so on....
Posted: Wed Apr 27, 2005 11:13 pm
by vmcburney
I am going to make a controversial statement here. I think in two years DataStage will have parallel jobs but it will not have server jobs. If you are starting out with DataStage you are better off going parallel.
Ascential are moving the entire product suite onto a parallel engine, they have put server jobs onto autopilot and they are developing and enhancing parallel jobs at a rapid rate. They are moving some of the best features of server jobs into parallel jobs, the transformer and the lookup stage are examples. And it now runs on Windows and Mainframe USS.