"distributed computing,p2p,grid,clusters,cuda?"

frito

Hi there

is there any distributed computing feature available in RM? General CUDA support? I could not find any info except this mighty old topic
http://rapid-i.com/rapidforum/index.php/topic,12.0.html

thx

(to admin: please remove the same topic in getting started section ,thx)

IngoRM

Hi,

beside the Parallel Extension which enables support for multiple cores for certain operators, the latest version of the Enterprise Edition of RapidAnalytics already offers cluster support. At last year's RCOMM, a first new extension using Map&Reduce / Hadoop was presented. Another extension for Hadoop offering even more algorithms will be presented at this years RCOMM so you might be interested in participating:

http://www.rcomm2011.org

From my point of view, CUDA is a completely different topic. However, there are several attemptions from different people worldwide who are currently in the process of adding new operators to RapidMiner which are based on CUDA. I am sure it is only a matter of time until the first extension will be released for CUDA support.

Cheers,
Ingo

frito

Hi Ingo,

[quote author=Ingo Mierswa]
Hi,

beside the Parallel Extension which enables support for multiple cores for certain operators, the latest version of the Enterprise Edition of RapidAnalytics already offers cluster support. [/quote]


Where can I read more about the cluster computing and its deployement in the RA Ent.?

Any trial available?

Is it planned for RA Community version too?

[quote author=Ingo Mierswa]
At last year's RCOMM, a first new extension using Map&Reduce / Hadoop was presented. Another extension for Hadoop offering even more algorithms will be presented at this years RCOMM so you might be interested in participating:

http://www.rcomm2011.org
[/quote]

Sounds tempting.

[quote author=Ingo Mierswa]
From my point of view, CUDA is a completely different topic. However, there are several attemptions from different people worldwide who are currently in the process of adding new operators to RapidMiner which are based on CUDA. I am sure it is only a matter of time until the first extension will be released for CUDA support.
[/quote]

I am not sure how feasible that would be but are you planning on some kind of a Cuda wrapper or an extra operator that one could add/encapsulate to the contemporary/non cuda Operators? Or something similar to your cluster computing solution utilizing RA?

[quote author=Ingo Mierswa]
Cheers,
Ingo
[/quote]

cheers
f

haddock

[size=10pt][size=10pt][size=10pt][size=10pt][size=10pt][size=10pt][size=10pt][size=10pt][size=10pt][size=10pt][size=10pt][size=10pt][size=10pt][size=10pt][size=10pt]What a load of crap.[/size][/size][/size][/size][/size][/size][/size][/size][/size][/size][/size][/size][/size][/size][/size]

IngoRM

Hi again,

Where can I read more about the cluster computing and its deployement in the RA Ent.? Any trial available?

Please contact our sales people for more information about the Enterprise Edition of RapidAnalytics. Trials are also possible.

Is it planned for RA Community version too?

Nope, sorry. There are a couple of things which are only for our Enterprise customers.

I am not sure how feasible that would be but are you planning on some kind of a Cuda wrapper or an extra operator that one could add/encapsulate to the contemporary/non cuda Operators? Or something similar to your cluster computing solution utilizing RA?

We are currently not planning anything on the CUDA front ourself. One of the groups currently working on that is the Artificial Intelligence Unit of the University of Dortmund, the old "home" of RapidMiner. First algorithms have been ported already as part of a masters thesis finished a couple of weeks ago. And since RapidMiner and the new library both have its source in Dortmund, I assume its only a matter of time until things get combined.    I am, however, no expert on this topic so I am not sure how the technical details would look like with respect to this combination.


@haddock
Pardon? I can hardly believe that somebody who posted almost 700 posts here in the forum actually answered like that...


Cheers,
Ingo

haddock

@haddock Pardon? I can hardly believe that somebody who posted almost 700 posts here in the forum actually answered like that...

I think it means that I am p***** off, not least because you cannot figure out why.