DSXchange

Could anyone suggest a good configuration of a UNIX server for installing Datastage? This will be an isolated Datastage server and no database or any other applications running on it.

Hello Ridar,

that is an impossible question to answer directly; in the car world that would be the equivalent of asking "I would like to buy a car, any recommendations?".

Here are a couple of questions whose answers will influence the decision:

- What flavor of UNIX and/or hardware is already in-house?
- How many concurrent users will design or concurrent production jobs run?
- Which components of DataStage are going to be used?
- What level of complexity are the jobs going to be?

This provides a good starting point for deciding architecture and sizing. Usually the realities of business will modify the original specifications for CPU(s) and memory downwards, since an ideal machine would just load everything into memory and use just a portion of it's processing and networking capabilities to run the jobs.

Hi Arnd,
Thanks for the response. We are now only going to do a new setup. So would like to know what are the best possible servers that we could use. Regarding the no. of users it would be around 5 and not real time production. It will be a kind of offshore development. Not a very heavy load initially, but, as time goes on and testing begins there might be some load testing also to be done. As far as components goes we will be using only Datastage(Administrator,Designer,Director & Manager). Complexity of jobs, well, cannot guess at this moment. But, might not be too simple.

ArndW wrote:Hello Ridar,

that is an impossible question to answer directly; in the car world that would be the equivalent of asking "I would like to buy a car, any recommendations?".

Here are a couple of questions whose answers will influence the decision:

- What flavor of UNIX and/or hardware is already in-house?
- How many concurrent users will design or concurrent production jobs run?
- Which components of DataStage are going to be used?
- What level of complexity are the jobs going to be?

This provides a good starting point for deciding architecture and sizing. Usually the realities of business will modify the original specifications for CPU(s) and memory downwards, since an ideal machine would just load everything into memory and use just a portion of it's processing and networking capabilities to run the jobs.

Continuing for Arnd....


How much data are you going to process? Designing an ETL application for moving thousands of rows of data can be a lot different from a system that moves millions of rows of data a day.

What kind of data? From a database to a database? From copy books to flat files? What are you doing?

The easiest answer is:

What is your DS license for platform and cpu count? One cpu Windoze machines are sufficient for a single developer to write DS jobs and test small amounts of data. But, to simulate and test scalability you need a larger machine. With 5 developers you will easily overpower a single cpu Windoze machine just using Designer and running small jobs.

So, your answer is that you need to give more information. To continue the analogy, you want to buy a car but you haven't told us how many people your car needs to transport, what kind of fuel efficiency you want, how much cargo space, how fast you want to go, and what price you are willing to pay.

ridar wrote: So would like to know what are the best possible servers that we could use.

the easy answer to this question is, the one that you can pay...