Home » forums » LtU Forum

In what case a system doesn't benefit from MapReduce? Do you know any example which doesn't benefit from MapReduce

I am making a research on MapReduce.

It's easy to find out about MapReduce. But my concern is about the systems that don't benefit from Mapreduce. Why they don't?

I want to search about a few softare systems which didn't/don't implement , benefit from MapReduce.

Any information will help.

Thank You

By boxinthebox at 2008-05-07 03:36 | LtU Forum | previous forum topic | next forum topic | other blogs | 5177 reads

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

primitive

1. This seems off-topic, and has been repeatedly asked about on this site.

2. While map/reduce may be sufficient to parallelize any problem, that does not imply it does so efficiently. The reduce step, essentially communication, does not immediately support locality (think an n-body simulations where only neighbors in some dimension need to share data) nor the mixing of communication and computation (it performs, largely, in lockstep).

Take a look at the Berkeley View technical report for problems deemed representative of critical parallelization challenges in general, let alone efficiently implementable using map/reduce in particular.

(EDIT: missed the word 'not' in bullet #2)

By lmeyerov at Wed, 2008-05-07 05:13 | login or register to post comments