Hi Emre I will not consider myself a experienced member but ill have a shot at this.
Mozart Oz has network transparency aimed at open networks, not just high performance clusters (erlang) or inter-core concurrency (clojure) though Mozart Oz can handle all of them quite well. The message passing actor based model is just one style of organizing your project. This can be emulated in Mozart Oz by creating what are known as (Mobile) Agents. Agents communicate with each via a mechanism called Ports. Which is basically a thread with a recursive procedure reading a dataflow stream and performing actions on the data then using a send and receive type operation. Now depending on your problem - you can chose to move the data, and keep the agent local, another way is to move the agent and keep the data local. It seems the map reduce computation model is becoming the new golden hammer, just because the paradigm works for Google's problem doesn't mean it is compatible for my little problem. I believe in comparison to Mozart Oz, swarm's network transparent algorithms - I am going to presume - are rather inefficient as the project is so new, Mozart Oz has already had these kinks knocked out of it*. Lastly the behaviour we have spoken about is _built into the language_ and not an included library. Also Mozart Oz has two garbage collection systems. One is local and the other global. Swarm has no garbage collection - since last I checked. It seems each swarm node needs to be homogenous whereas this isn't the case for Mozart Oz (though it could be) I hope my answer has done justice to your question. Stewart *except a couple On Thu, Jul 28, 2011 at 12:30 AM, Emre Sevinc <[email protected]> wrote: > Hello, > > I know that Mozart-Oz lets me run computations in network-transparent > manner. I wonder if that is similar to what Scala achieves with actors > and continuations, especially in the context of a project such as like > Scala Swarm: > > http://code.google.com/p/swarm-dpl/ > > "Swarm is a framework allowing the creation of web applications > which can scale transparently through a novel portable > continuation-based approach. Like Map-Reduce, Swarm follows the maxim > "move the computation, not the data". However Swarm takes the concept > much further, allowing it to be applied to almost any computation, not > just those that can be broken down into map and reduce operations." > > I would be happy if experienced members of this list can elaborate on > this topic and compare the approaches. > > Kind regards, > > -- > Emre Sevinç > _________________________________________________________________________________ > mozart-users mailing list > [email protected] > http://www.mozart-oz.org/mailman/listinfo/mozart-users > _________________________________________________________________________________ mozart-users mailing list [email protected] http://www.mozart-oz.org/mailman/listinfo/mozart-users
