server stages vs parallel stages

agpt · Post by **agpt** » Wed Sep 29, 2010 1:49 am

Why it is so that we do have some stages in parallel but not in server? ?Parallel jobs are meant to use benefits of parallelism but and running them as server job might not give that benefit. Bu still why they are not available at all?

chulett · Post by **chulett** » Wed Sep 29, 2010 6:24 am

Because... they're not? Seriously, not really sure what kind of an answer you were expecting. The two products have completely different origins, philosophies and architectures so the vast majority of the stages specific to PX wouldn't really apply in the Server world. Common ones do, of course, like sort and aggregator.

Even if they could be, IBM won't really be investing much (if any) new R&D monies in the Server product as PX is their tool for the future. Out of curiousity, did you have an example or two in mind?

agpt · Post by **agpt** » Wed Sep 29, 2010 10:25 am

chulett wrote:Because... they're not? Seriously, not really sure what kind of an answer you were expecting. The two products have completely different origins, philosophies and architectures so the vast majority of the stages specific to PX wouldn't really apply in the Server world. Common ones do, of course, like sort and aggregator.

Even if they could be, IBM won't really be investing much (if any) new R&D monies in the Server product as PX is their tool for the future. Out of curiousity, did you have an example or two in mind?

,

Craig,

So how do we decide whether to implement server or parallel job for example say for sorting? Does it depend on the whether i want to use parallelism facility or not? and if this is the case that IBM won't be spending R&D money at 2 places, why do we have even few common jobs?

For an example why can't we have SCD or compress stage in server ?

PX is tool for future? is it mean Server is going to be obsolete?

ray.wurlod · Post by **ray.wurlod** » Wed Sep 29, 2010 5:14 pm

Not obsolete. Some functionality (such as Connectors) will find its way into server jobs. But the server job architecture does not warrant the "one task, one stage type" functionality of parallel jobs. Rather, server jobs tend to use the Transformer stage as the workhorse.