Why use server job?

taral · Post by **taral** » Tue Jul 20, 2010 5:52 am

why use server job when we have parallel jobs which run much faster than server jobs?

eostic · Post by **eostic** » Tue Jul 20, 2010 6:05 am

...do some searches...this has been deeply covered in prior threads.

ArndW · Post by **ArndW** » Tue Jul 20, 2010 6:10 am

Your assumption that parallel jobs [always] run faster than server jobs is flawed.

dr.murthy · Post by **dr.murthy** » Tue Jul 20, 2010 7:31 am

Just simply do a sort test with some huge data on both server and parallel jobs , then you can easily idnetifie which works faster

priyadarshikunal · Post by **priyadarshikunal** » Tue Jul 20, 2010 7:35 am

dr.murthy wrote:Just simply do a sort test with some huge data on both server and parallel jobs , then you can easily idnetifie which works faster

Not a correct test to define benchmark. They both have different benifits and should be used to according to requirements to properly utilize the capabilities of the tool. And I strongly agree to Arnd's comment.

vivekgadwal · Post by **vivekgadwal** » Tue Jul 20, 2010 7:39 am

priyadarshikunal wrote: Not a correct test to define benchmark. They both have different benifits and should be used to according to requirements to properly utilize the capabilities of the tool. And I strongly agree to Arnd's comment.

Agreed!

One of the things that works in favor of server jobs is the job start-up time. Also, Vincent McBurney has some very good posts in ITToolBox website with regards to this. Please do some reading in here as well as in the above mentioned site so that you can form an understanding for yourself.

taral · Post by **taral** » Tue Jul 20, 2010 10:05 pm

what is the name of the site?

chulett · Post by **chulett** » Tue Jul 20, 2010 10:09 pm

ITToolBox. Name and URL both, oddly enough.

ray.wurlod · Post by **ray.wurlod** » Wed Jul 21, 2010 12:47 am

http://it.toolbox.com

chulett · Post by **chulett** » Wed Jul 21, 2010 5:50 am

www.ittoolbox.com auto-redirects, hence my comment.

ray.wurlod · Post by **ray.wurlod** » Wed Jul 21, 2010 6:04 am

Why would you create a parallel job to process a single row?

Since parallel jobs move data in chunks of not less than 32KiB, and typically much larger slabs (128KiB, 1MiB or 3MiB), why would you use a parallel job to process less then 32KiB of data?

taral · Post by **taral** » Thu Jul 22, 2010 12:13 am

Did not find any blog related to server job performance and in which condition to use server jobs.

ArndW · Post by **ArndW** » Thu Jul 22, 2010 4:58 am

I don't know of any blogs on this rather esoteric subject, but if you search DSXChange you will find many threads that touch upon the differences between Server and Parallel job. By looking at the relevant strengths and weaknesses of the two types of jobs one can easily think of scenarios where one or the other is much better either in terms of design simplicity or runtime efficiency or resource usage.

vivekgadwal · Post by **vivekgadwal** » Thu Jul 22, 2010 8:14 am

taral wrote:Did not find any blog related to server job performance and in which condition to use server jobs.

You did not look hard enough then!

For starters:
http://it.toolbox.com/blogs/infosphere/ ... stats-6976

priyadarshikunal · Post by **priyadarshikunal** » Thu Jul 22, 2010 9:29 am

that blog outlines the benifits of parallel jobs oves server. But there are some benifits of using server jobs over parallel. Just to mention a few:

1. its faster with very low volume of data, also no startup time.
2. works really well with flats files and doesn't even complains about few dodgy stuffs you generally find in flat files.

Too much to write, but this should give you a start.