Configuration File

Bicchu · Post by **Bicchu** » Tue Oct 18, 2011 9:06 am

Why in a configuration files nodes are always present in a multiple of two?

Thanks in advance.

jwiles · Post by **jwiles** » Tue Oct 18, 2011 10:25 am

Human nature more than any other reason...there is no software or hardware requirement for it.

Regards,

Bicchu · Post by **Bicchu** » Tue Oct 18, 2011 11:19 am

Is that so? Or there is any logical reason behind this?

jwiles · Post by **jwiles** » Tue Oct 18, 2011 1:25 pm

Once again, human nature or force of habit. It's quite natural to think in terms of multiples of two: 1, 2, 4, 8 and so on, or in even numbers: 2, 4, 6, 8, etc. Remember, the digital computer world is based on binary logic: powers of two.

Or, from another point of view: When adding nodes to an existing configuration file to increase the parallelism of your job, would you be more likely to a) Double the number of nodes, b) Triple the number of nodes, or c) Add one more node? It's usually much easier for most people to just copy and paste the entire set of nodes (i.e. double the nodes) when editing the config file.

DataStage doesn't care...it will happily (well, as happily as software can) attempt to start your job with however many nodes are defined, whether it be a multiple of two or twenty two, even or odd, or 42, so long as the config file is valid.

Regards,

ray.wurlod · Post by **ray.wurlod** » Tue Oct 18, 2011 4:04 pm

One thing that gives me the "irrits" is those who suggest running on a single node as the way to "solve" poor partitioning choices.

It's not a solution.