Hello,
Here is an interesting paper of Google's MapReduce reverse engineered
into Haskell. I apologize if already posted .
http://www.cs.vu.nl/~ralf/MapReduce/
Kind regards, Vasili
___
Haskell-Cafe mailing list
Haskell-Cafe@haskell.org
http://w
Any idea how one might implement a multi note map and reduce network? Lets
assume I have network of nodes that act as master and salve. How can I take
a user code (containing his map reduce logic) and actually run it on
different nodes?
Daryoush
2009/2/24 Galchin, Vasili
> Hello,
>
> He
Vasili,
What do you mean? Googles MapReduce is already a published / well
understood concept so no reverse engineering is needed. If you are
asking about pre-existing implementations, there is at least one [1]
but only for reference, not speed. If you are asking about community
interest, great a
I've read this article too, and I must say that it is indeed a very
interesting and exciting read, both in terms of understanding
MapReduce and its capabilities somewhat better, and in terms of
beholding the beauty of Haskell.
It is not exactly reverse engineering, but it is expressing the
essense
The author(s) are from Microsoft .. they claimed to reverse engineer
internals. My understanding is that MapReduce internals **is** a Google
closely held secret ... hence the open source version is "dumbed down" .. I
have forgotten the name.
Vasili
On Wed, Feb 25, 2009 at 8:49 AM, Eugene Kirpicho
Galchin,
Maybe you are asking not only about remote execution, but also mobility of
code. This is a problem that is previous to mapReduce, since mapReduce
assumes that all the code (and the data) is in place in the respective
nodes. In fact, the distribution of resources in order to efficiently us
I agree. A distributed database could be made as usable as a standard
RDBMS by offering an interface by which you supply a map/reduce pair
of functions and a list (range?) of keys.
This could be easily implemented with a database such as Scalaris, in
which the Chord algorithm is responsible for pl
2009/2/25 Alberto G. Corona
> And it would solve a lot of problem: scalability, system re-configuraition
> and installation: just by adding or removing nodes at runtime.. heavy
> numerical computations are also good candidates.
>
> 2009/2/25 Rick R
>
> I agree. A distributed database could be
>
> since mapReduce assumes that all the code (and the data) is in place in the
> respective nodes.
>
As far as I can tell, the Hadoop, the open source implementation of map
reduce, doesn't require your map reduce code to be in all nodes. It copies
the jar files of the your application to the no