help for parallel search

[Lucia Cambise]

I am studying Oz using the documentation and, in particular,
I am using a simple problem (Magic Square) to learn Constraint Programming
in Oz and how a problem
can be solved using multiple networked computers.

I fully understood the problem and Constraint Programming on a single
computer, but I still have some problems
in solving the same problem on a network of computers.

I have some doubts, so I have some questions. I hope someone could help me
finding answers.

1) In the section "Parallel Search Engines"  of the documentation I read
that there is a method "trace(+B)", that
switches graphical tracing of search tree delegation on or off, depending on
the value of B.
I tried this method to solve the problem of Magic Square using multiple
networked computers. I expected to
see a window like that of the Explorer, that could help me to verify how the
search tree is explored and how the
subtrees are delegated to other nodes on the network. Instead, the method
opens a window that shows how many search tree nodes a
computer explored. Is there anyone who could tell if there is a method to
control the search real time (like the Explorer)
using a parallel search engine? Thanks.

2) I am using ssh as a fork method. After the connection between computers
on the network is established, do You know if
the computers, during the search, communicate through ssh or some ports? I
make this question because the ssh on the computers
I have to use is based on Kerberos, so there is a limit of time in
authentication. For example, if the time expires, I lose the authentication,
and I don't know if the search continues or if an error occurrs. So, if
computers communicate through ssh during all the time of the search
I have to authenticate again. On the contrary, if computers communicate
through some ports, the re-authentication I think is not necessary.
I don't know if  I am clear in explaining this problem.

3) Last question: does anyone know how the the computers on a network divide
the load and how the problem is split between networked computers?


Any help in my learning would be much appreciated. Thanks in advance for all
the answers.

Lucia
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.gforge.info.ucl.ac.be/pipermail/mozart-hackers/attachments/20060215/2412eab9/attachment.html