Distributed Functionnal Object Computations

The OASIS INRIA Sophia Antipolis project has been designing and implementing  ProActive, a Java library for Parallel, Distributed, and Concurrent computing and metacomputing. In the framework of a MIMD model (active object model), from a reduced set of rather simple primitives, a comprehensive and versatile library is defined for parallel, distributed, and concurrent programming (middleware).
In the absence of any syntactical extension, ProActive PDC programmers write standard code. The library is itself extensible by the programmers, making the system open for adaptations and optimizations.

The University of Adelaide team of Andrew L Wendelborn (http://www.cs.adelaide.edu.au/users/andrew/) has been working on Parallel functional programming, and Process Network implementation within  MetaComputing environements.

Whithin this framework, this Researsh Master offers to tackle the following challenges:

• Design a functionnal subset of Java (called JavaF thereafter) that is suitable for distributed  and functionnal object computations
• Implement the subset within the ProActive library
• Works with IC2D for visualization and control
• Achieve meta-computing experiments with ProActive, IC2D, and the JavaF subset
• Study how it relates to the Process Network ideas

• Non-mutable subset:
• The order of instructions is not significant; only data dependancies are
• Single assignment
• Both constructors and functionnal operations:no procedure calls, only functions
• Asynchronous constructions of objects and function calls
• Copy semantics (deep) for calls to JavaF objects
•  Wait-by-necessity
• Mutable subset:
• Procedure calls are allowed: the way to do side-effect operations on objects
• The order of instructions become significant
• Still have the following:
•    Asynchronous constructions and function calls
•    Copy semantics (deep) for calls to JavaF objects
•    Wait-by-necessity
• Similar features for procedure calls.

One strong advantage of Process Network to take into account being: that framework guaranties determinate, deadlock-free applications.
It might also be usefull to study the notion of future-based streaming in general, and or the use of dedicated buffered demand-driven computations
in order to achieve implementation that fit within finite-size buffers.

Both standard and functional Java should be able to exist in same application; this calls for a study in order to precisely define the boundaries of the
functional part.

These researsh should lead to a PhD program (Thèse de Doctorat).

Encadrement :  Françoise Baude; Denis Caromel; Isabelle Attali;
Téléphone : 04 92 38 76 71 Email : fbaude@sophia.inria.fr,Denis.Caromel@inria.fr
Laboratoire ou équipe : INRIA Sophia Antipolis -- I3S CNRS
 

En collaboration avec :  Dr Andrew L Wendelborn, Department of Computer Science, University of Adelaide, Australia
Téléphone : +61-8-8303-4726 Email :andrew@cs.adelaide.edu.au