Page 1 of 1

Diffrence on Datastage MVS edition and Parallel extender

Posted: Wed Jul 16, 2008 12:40 am
by mohandl
Hi

We are doing POC work for our company (migration from IMS to db2).We are assuming two Approches 1. From Parallal extender 2.from Datastage MVS Edition,can you please let me konw diffrence on following points

1. Performance / Scalability of data which one is bettter MVS or Parallel ?
2. Ease of developement which one is bettter MVS or Parallel ?
3. Maintanability which one is bettter MVS or Parallel ?
4. Cost which one is bettter MVS or Parallel ?
5. Ease of deployment which one is bettter MVS or Parallel ?

Posted: Wed Jul 16, 2008 1:22 am
by ray.wurlod
1. A definite maybe. Depends on configurations.
2. MVS. Fewer stages to know/learn. Less flexibility as a result.
3. Equal. "Will IBM continue to support MVS edition?" might be a pertinent question to ask.
4. Define "better". Either can be more expensive, depending on the configuration.
5. MVS, if you ignore the politics of getting permission to run stuff (generated COBOL and JCL) on the mainframe.

And, just to confuse matters further, there is also Enterprise Edition on USS (a partition of your mainframe running UNIX). Parallel jobs can get almost "native" access to mainframe-based database tables and files.

Posted: Wed Jul 16, 2008 2:22 am
by vmcburney
Cost is a complex question and depends on the architecture you are interested in. I'm assuming both databases reside on the mainframe and you don't want to pull the data off, transform it and push it back on again. Without a doubt the cheapest way to use DataStage on the mainframe is Information Server for z/Linux. It's daylight to second.

Consider the MVS approach - huge up front license fees. Something in the magnitude of 800K. Then you pay forever for MIPS - mainframe processing that DataStage jobs will chew through. Plus you pay for an off mainframe development environment - not much in the way of license fees but you have to pay to maintain the windows/unix/linux server.

z/Linux is a partition on your mainframe with a special version of Linux that can run Information Server and Cognos software. It's priced per CPU rather than by the MIPs and it's cheaper to setup and run than MVS.

There is also a DataStage for Unix System Services that lets you run jobs in a Unix partition on the mainframe.

Both the USS DataStage and z/Linux DataStage use the standard parallel job canvas so it's familiar to DataStage developers. MVS has it's own set of stages and job designs so it's harder to find a specialist. So z/Linux beats MVS for cost and maintainability. MVS may be faster as it runs in native mainframe but that's not a big consideration in a conversion.