[ndnSIM] Distributed Application in CCN domain

[Ilya Moiseenko]

1) Usually, multi-threaded programs have some thread model - either a thread pool, or thread creation/destruction while keeping the current number of active threads.
My point is that there is a shared data structure to keep references to the threads, and you sort of know what each thread is doing.

In case there are multiple thread-pools in your process, your naming model can look like this:
/localhost/chatapp/workerPool/thread1
/localhost/chatapp/workerPool/thread2

/localhost/chatapp/managerPool/thread1
/localhost/chatapp/managerPool/thread2

and so on. And you can develop a forwarding strategy that will do broadcasting for example to /localhost/chatapp/workerPool  (for all threads in this pool), or multicasting with retransmission, etc.

2) Yeah, it’s MPI related. Are you sure that you need ndnSIM for testing inter-process communication inside the localhost?  I would use NFD.


On Sep 30, 2014, at 11:47 PM, Haroon Rashid <haroonr at iiitd.ac.in> wrote:

> Some more clarifications:
> 
> 1>  By CCN I mean CCN/NDN paradigm and not the CCN daemon. How come a thread would come to know the name of another thread, because threads created by a parent process are unaware of each other.
> 
> 2>  Right now my concern is only about ndnSIM. I don’t know anything about NFD. While searching, I found there is something MPI in NS-3, but I am not sure do I need this one or something else.
> 
> 3>  By distributed application I mean an application which can be divided into some threads such that each thread work parallely. End result of such a process execution should be same as that of its serial execution. 
> 
> 
> On Wed, Oct 1, 2014 at 5:15 AM, Ilya Moiseenko <iliamo at ucla.edu> wrote:
> I don’t work with CCNx daemon, but with NFD daemon each of your threads can register its own individual name like
> 1) /chatapp/thread1
> 2) /chatapp/thread2
> 3) /chatapp/thread3
> 
> I didn’t understand question b), but you can have intra-process communication with Interest and Data packets in NFD.
> 
> Regarding ndnSIM, in normal mode, its core works in a single thread, and it’s not that trivial to make your simulation scenario run in multiple-threads. 
> You can refer to general NS-3 documentation for help.
> 
> To find some information about “pushing”  you can find the section talking about "Command Interests" in NFD developers guide.
> 
> Ilya
> 
> On Sep 30, 2014, at 4:32 AM, Haroon Rashid <haroonr at iiitd.ac.in> wrote:
> 
>> Dear All,
>> 
>> I am curious to know how a distributed application works in CCN paradigm. Consider a scenario where an application is having three threads (logical processes) running on three nodes (three processors of a single node). Now, if these three processes want to communicate with each other for a certain reason. How can I achieve the same in CCN, i.e.,  (a) how processes can refer to each other (NAMING CONCERN) (b) will the format of Interest and Data packets will be same as that of current packet structure in CCN/NDN. Since, these communication messages might be very short like Hello packets or status request of a node.
>> 
>> 
>> 
>> Is there any simple distributed application with which I can start working and come out with different possibilities of communication between processes of an application? Furthermore, I would like to mention that I want do this in ndnSIM.  Currently, is there any support (libraries) in ndnSIM for doing this.
>> 
>> 
>> 
>> Another concern I am thinking is that  I know that CCN is pull based approach. Can I mix a push based approach for some messages in a distributed scenario discussed above without violating CCN  paradigm?
>> 
>> 
>> 
>> Best, 
>> 
>> Haroon Rashid
>> 
>> 
>> _______________________________________________
>> ndnSIM mailing list
>> ndnSIM at lists.cs.ucla.edu
>> http://www.lists.cs.ucla.edu/mailman/listinfo/ndnsim
> 
> 
> 
> 
> -- 
> Haroon Rashid
> 
> 
> _______________________________________________
> ndnSIM mailing list
> ndnSIM at lists.cs.ucla.edu
> http://www.lists.cs.ucla.edu/mailman/listinfo/ndnsim

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.lists.cs.ucla.edu/pipermail/ndnsim/attachments/20141001/13c4129c/attachment.html>