Moving from DataStage Server to Enterprise Edition

[ianst]

Hi,

I would be grateful if you could share your experiances of moving from using standard datastage server to enterprise edition?

What is the learning curve?

Are the two products totally different?

What extra skills are needed to use it?

Thinks like that.

We are currently in the situation when for performance reasons we are considering making the move.

Thanks for your thoughts.

Ian

[legendkiller]

Hi,

parallel extender has many more functinality as comapred to server addition, but also we have take into consideration many things also.

specially we have to concentrate on configuration files and partition funda which are key in achiving desired functionality in parallel extender

regards
legendkiller

[ray.wurlod]

It is possible, with a number of severe restrictions, to use server job components in parallel jobs. Read about these in the Parallel Job Developer's Guide (the size of this manual will give you a strong hint about the learning curve).

The two products ARE totally different, with only a small overlap. However, the design metaphor (stages and links) remains.

The biggest differences you have to get you head around are that hashed files are no longer available (there are three different ways to do lookups), and you have to come to grips with the idea of partitioning, including specifying processing nodes in a configuration file. There are many more stage types than in server jobs, the stages have highly specific tasks and you need to develop the skill of choosing the right stage type for the particular task. Just memorize the manual!

[T42]

What is the learning curve?

You must go away from the Transformer mentality to a wide compliment of stages that handles the same functions (lookup stage for one.)

Basically, it requires flexibility on the person learning about the product to handle different concepts. It's like learning Java after learning C++. Appears the same, but totally different.

What extra skills are needed to use it?

Patience. Lots of patience.

Be careful on pre-established assumptions when using EE.

[ianst]

Thanks for your responses. It sounds that if we go for enterprise edition, that we will have to assume that we are starting from scratch and allow time accordingly.

[T42]

Thanks for your responses. It sounds that if we go for enterprise edition, that we will have to assume that we are starting from scratch and allow time accordingly.

Not necessarily. You could still run Server jobs. Just do a phased switch -- focus on certain jobs that are not performing to your preferences, and switch it to PX using standard input/output (not datasets or hash files). Test, and deploy. Keep phasing in as much as you reasonably can.

[vmcburney]

While the two job palettes have a different set of stages in them the administration around executing those jobs and checking the results is still largely the same. Both can be started by the Director, batch jobs, sequence jobs or job control routines. So you can run with a mix of both types and as T42 recommended just rebuild those that have performance problems.

When you install the software make sure you fill in both the standard and enterprise authorisation keys so you get both sets of job stages.