looks like a peer-to-peer network to me. here are some libs:

https://github.com/Raynos/peer-connection-network

or you can develop it as a grid (read p2p network with predefined peers) 
using https://github.com/substack/dnode

Am Montag, 18. Februar 2013 18:33:47 UTC+1 schrieb MikeB_2012:
>
> Hi all.  I'm an experienced data mining/machine learning algorithm 
> developer redirecting from programming/prototyping in Matlab (past decade) 
> to programming in js/node.js. I'm looking for guidance as to what node 
> packages I should be considering to achieve a certain functionality.  
>
> The functionality I'm looking for is approximately as follows:
>
>   a)  an assortment of sensors puts data on a network.  Each sensor has a 
> dedicated process that pre-processes sensor data.  The processed data is 
> then transmitted to other processes;
>   b)  there are several virtual 'layers' or 'groups'.  Each layer has a 
> specific purpose or goal that is attained through the asynchronous 
> processing of the processed sensor data by one or more processes;
>   c)  should any process fail/crash, the remaining processes are 
> unaffected;
>   d)  similarly, processes can be added to the system as required without 
> affecting other processes; and
>   e)  processes take input from other processes.  For example, some 
> processes obviously will use the output of sensor pre-processing 
> processes.  But some processes will simply be taking the output of other 
> processes.
>
> So you can picture this network of processes (a 'mesh' ?) with some 
> subsets of the mesh working towards common goals.  The idea is vaguely 
> comparable to neural networks but with large amounts of feedback say, a 
> liquid state NN vice multi-layer feed forward NN, complex processes vice 
> simple transfer functions at the mesh nodes, and asynchrony.
> Within the mesh, processes are born and die, but the whole operation 
> remains largely unaffected in the sense that other processes will only 
> lose/gain information but they won't fail (cease trying to process).  The 
> latter is important because a nuance is that a given process will decide 
> for itself what information it will use, what other processes it will 
> 'listen' to and will have to adapt when changes occur.  So processes must 
> be aware of each other.
>
> Anyways, I can fake portions of all this using Matlab (in un-real-time) 
> but node.js seems to offer the potential to actually implement the idea in 
> real-time, kind of, maybe.  (I'm still learning so some of you may know 
> better.)  I was optimistic with one node package, hook.io, but it has 
> shut down.  I'm happy to fork it but would need lots of help and guidance.  
> Several others were suggested including: kue, rabbitmq, zero-mq, nssocket, 
> dnode, amino, and federation. But whereas I'd found a clear explanation 
> of hook.io (with illustrations to confirm it seemed like what I 
> wanted)<http://ejeklint.github.com/2011/09/23/hook.io-for-dummies-part-1-overview/>most
>  other packages need knowledge (eg. - reddis) that I don't yet have 
> ('federation' 
> being the exception <https://github.com/jacobgroundwater/federation>).
>
> So that's what I'm trying to do and it looks like js/node.js can help me 
> do it.  Is that a correct conclusion?  If so, are there any packages or 
> approaches that you might suggest?  As always, any constructive guidance or 
> advice would be appreciated.
>

-- 
-- 
Job Board: http://jobs.nodejs.org/
Posting guidelines: 
https://github.com/joyent/node/wiki/Mailing-List-Posting-Guidelines
You received this message because you are subscribed to the Google
Groups "nodejs" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to
[email protected]
For more options, visit this group at
http://groups.google.com/group/nodejs?hl=en?hl=en

--- 
You received this message because you are subscribed to the Google Groups 
"nodejs" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/groups/opt_out.


Reply via email to