ISPDC 2014


13th International Symposium on Parallel and Distributed Computing

Aix-Marseille University - Porquerolles Golden Island


 Gul AGHA, University of Illinois, USA

Actors Programming for the Mobile Cloud

Actor programming languages
provide the kind of inherent parallelism that is needed for building applications in the mobile cloud. This is because the Actor model provides encapsulation (isolation of local state), fair scheduling, location transparency, and locality of reference.  These properties facilitate building secure, scalable systems. 
Not surprisingly, very large-scale applications such as Facebook chat service and Twitter have been written in actor languages. 

The lecture will introduce the basics of the actor model and then discuss the state of the art in actor language theory and practice.  In particular, I will describe the status and open problems in reasoning systems for actors, performance and energy efficiency issues related to their implementation, and the challenge of facilitating secure coordination between actors.

Bio: Gul Agha is Professor of Computer Science, and of Electrical and Computer Engineering, at the University of Illinois at Urbana-Champaign.  Dr. Agha's research is in the area of programming models and languages for open distributed and embedded computation. Dr. Agha is a Fellow of the IEEE. He served as Editor-in-Chief of IEEE Concurrency: Parallel, Distributed and Mobile Computing (1994-98), and of ACM Computing Surveys (1999-2007). 
He has published over 200 research articles and supervised over 20 PhD dissertations.  His book on Actors, published by MIT Press, is among the most widely cited works.  Besides work on semantics and implementation of actor languages, Agha has done pioneering research in concolic testing, predictive monitoring,
computation learning for verification, statistical model checking and Euclidean model checking, and energy complexity of parallel algorithms.

Susanne GRAF, Verimag-CNRS Grenoble, F

Knowledge-based derivation of distributed implementations from global specifications


Deriving distributed implementations from global specifications has been quite extensively studied for different application domains, under different assumptions and constraints. The solutions that have been proposed are often relatively adhoc and developed independently for different application domains.
We explore here the knowledge perspective to achieve a potential for a more unifying approach: a process decides to take a local action when it has the knowledge to do so.
We discuss typical knowledge atoms that are useful for expressing local enabling conditions with respect to different notions of correctness, as well as different means for obtaining knowledge and for representing it locally in an efficient manner. Our goal is to use such a knowledge-based representation of the distribution problem for either deriving distributed implementations automatically from global specifications on which some constraint is enforced, or for improving the efficiency of existing protocols by exploiting local knowledge.

We also argue that such a knowledge-based presentation helps achieving the necessary correctness proofs.

Dan C. MARINESCU, University of Central Florida, USA

Parallelism: Memories of Time Past and a Glimpse at the Future

K.V.S. PRASAD, Univ. of Gothenburg, CHALMERS, SE

The evolution of broadcast models and calculi


The ALOHANET and then the Ethernet enabled broadcast as a communication primitive for concurrent or parallel programming. Speakers choose what to say and when, and are heard instantaneously by everyone connected to the (local) Ethernet, which also arbitrarily resolves any contention among speakers.  The first process calculus to use this model, CBS, appeared in 1991.  Later versions proved easy to implement and to use for concurrent and parallel programming, so that such calculi embedded in proof checkers such as Coq or Agda give correct executable concurrent programs with data.  Broadcast can easily accommodate  priorities and time, and can be encoded in CCS, but only by adding coordinator processes.  To local broadcast, later calculi added  mobility and asynchronous connections between locally synchronous nets.  Recent wireless calculi capture  broadcasts with limited and overlapping ranges, and are meant for applications such as MANETS and sensor networks.

Roger SHEPHERD, Microsystems Consultant, ex-Inmos-STMicroelectronics, Bristol UK

Parallel systems from 1979 to 2014: 35 years of progress?


35 yeas ago I went to work for Inmos, a semiconductor company which believed it could integrate a processor, memory and a communication system on to a single chip, and that the resulting device could become a new building block for electronic systems. Inmos launched such a device, the transputer, five years later in 1984.
In 2014 it would be possible to integrate about 10,000 transputers on to one chip but the world is not building its electronic systems out of massively parallel processor systems.
This talk will look at the factors which have caused this apparent lack of progress in the design of parallel systems.



Institut de Mathématiques
de Marseille-ERISCS


logo AMU
© 2013-2014 IML-CNRS - updated: december 13, 2013