DSXchange

What exactly is the difference between Parallel and Server editions of Datastage? I know the basic difference is that PX will use parallel and pipeline processing, but still I see some jobs are preferrable coded as Server jobs only.

Based on what parameters, are Server jobs preferred over Parallel jobs?

Please refer to the datastage doccumentation to get the diffrences b/w server and parallel jobs.
The data processing is sequential in server jobs.Server jobs need lesser startup time as compared to Parallel jobs and perform well for small amounts of data.Most of the transformations need to be done in the transformer stage.
The parallel jobs can process large amounts of data in lesser time when compared with server jobs bc of the parallel processing capabilities and the start up time is more as Datastage has to devide the available resources(prcessors) as nodes for attaining the parallelism.
There are many parameters like volume of the data to be processed, number of physical processors available on the server, buffer sizes,..etc to decide which type of jobs are better(server/parallel).

in_finity307 wrote:What exactly is the difference between Parallel and Server editions of Datastage? I know the basic difference is that PX will use parallel and pipeline processing, but still I see some jobs are preferrable coded as Server jobs only.

Based on what parameters, are Server jobs preferred over Parallel jobs?

Hi,

Most of the time server job's are used in in sequencer's
Like Job Intailization and Job end to have note on which time the job has been started and ended (it Seems that people doesn't want run small work on parllel... )

If u want more information start analyzing from your project where it used and reply the post.. anlyzed report..

Regards..Naveen.K

Generally PX jobs run with more throughput but take more time to develop. Let's take an example job that takes 3 days to develop in PX and runs in 10 minutes and the same job would take 1 days to develop in Server but runs for 60 minutes.

Which is better? If the job runs 10x a day and is part of a critical path then the answer is obvious. But what if it only runs once a week? Or runs several times daily but nothing depends on it? Do you have PX or Server expertise in house or need to hire external people? Who will maintain it? Will the job ever become a bottleneck?

Sometimes smaller data volumes will run faster in a Server job than in a Parallel one. Sometimes a Server job will run longer but, in total, consume less resources than a quicker running PX job.

As you can see, the answer is not straightforward and those installations which have rules like "no server jobs" are not using the tool to it's potential.

Jasti wrote:The data processing is sequential in server jobs.Server jobs need lesser startup time as compared to Parallel jobs and perform well for small amounts of data.

Sorry, just had to pop in here as a representative of the Luddite population who do not have access to PX jobs where they work. The first statement quoted above is untrue. And while the second technically is true, they can perform well for fairly large amounts of data as well. As with any product, it helps to understand how to design for the strengths of the product, what it takes to make it sing. And mine give 5 performances a week on the Fhloston Paradise.

I too am currently working at a server-only site, and parallelism is possible. Even without it, some jobs are processing tens of millions of rows quite efficiently.

The difference is in how the job types are executed. Parallel jobs are executed in a parallel execution environment, and so are suited for multiple machines, in cluster or grid environments. However server jobs can take full advantage of a multi-CPU machine.

Server jobs will be around (and supported by IBM) for a long, long time.

But most of the "new toys" will be available only for parallel jobs.