Section: New Results

Programming Languages for Distributed Systems

Multi-active Objects

Participants : Ludovic Henrio, Justine Rochas, Vincenzo Mastandrea.

The active object programming model is particularly adapted to easily program distributed objects: it separates objects into several activities, each manipulated by a single thread, preventing data races. However, this programming model has its limitations in terms of expressiveness – risk of deadlocks – and of efficiency on multicore machines. We proposed to extend active objects with local multi-threading. We rely on declarative annotations for expressing potential concurrency between requests, allowing easy and high-level expression of concurrency. This year we realized the following:

• We proved the correctness of our compiler from ABS language into ProActive multi-active objects. This translation can be generalised to many other active object languages. This work has been published as a research report, and is under submission to a conference. The proof brought us very deep and interesting understanding on the differences between the languages.

• We started to work on static detection of deadlocks for multi-active object. This is the work of Vincenzo Mastandrea who is starting a Labex PhD in collaboration with the FOCUS EPI (Univ of Bologna). An article is currently submitted to a conference on this subject.

• We are formalising in Isabelle/HOL a first version of the semantics of multiactive objects. This work was done in collaboration with Florian Kammuller

• We organised a workshop on active object languages with the main teams in Europe involved in the development of active-object languages. A journal survey paper on the subject is currently being written.

• We implemented a debugger for multi active object programs.

We plan to continue to improve the model, especially about compile-time checking of annotations and about fault tolerance of multiactive objects.

Behavioural Semantics

Participants : Ludovic Henrio, Eric Madelaine, Min Zhang, Siqi Li.

We are conducting a large study on Parameterised Networks of Automata (pNets) from a theoretical perspective. We started last year with some 'pragmatic' expressiveness of the pNets formalism, showing how to express a wide range of classical constructs of (value-passing) process calculi, but also complex interaction patterns used in modern distributed systems. After publishing those results [13] , we focused on open systems and our formalism is able to represent operators of composition of processes, they are represented as hierarchically composed automata with holes and parameters. We defined a semantics for open pNets and a bisimulation theory for them. This study was driven by several usecase examples including a hierarchical broadcast algorithm and several operators of concurrent processes. A short presentation is accepted for publication in the journal “Science China: Information Sciences -”. A full paper on the subject of the semantics and bisimulation for open pNets is under submission to a conference.

In parallel we have started the study of a denotational semantics for open pNets, based on the Universal Theory of Processes (UTP). The idea in the long term would be to draw links between the operational, denotational, and algebraic models of the pNet formalism. A short presentation of our preliminary results will be presented at the conference PDP'16 (work in progress session).

GPU-based High Performance Computing for finance

Participants : Michael Benguigui, Françoise Baude.

We have pursued our work on pricing American multi-dimensional (so very computation intensive) options in finance and we hav been able to extend this to the computation of Value At Risk (consists in repeating the American option pricing, but we have found a financial grounded optimization that avoids us to replicate the most time consuming phase).

Moreover, the balancing of work is taking in consideration the heterogeneous nature of the involved GPUs, and is capable to harness the computing power of multi-core CPUs that also support running OpenCL codes. As our scheduling solution is capable to get a reasonable prediction of the workload of each slave computation, we have leveraged this to run the whole pricing and VaR computations onhybrid and heterogeneous clusters. These last results have been incorporated in the PhD thesis of M. Benguigui.

Scalable and robust Middleware for distributed event based computing

Participants : Maeva Antoine, Fabrice Huet, Françoise Baude.

In the context of the FP7 STREP PLAY and French SocEDA ANR research projects terminated late 2013, we initiated and pursued the design and development of the Event Cloud.

As a distributed system handling huge amount of information, this middleware can suffer from data imbalances. In a journal extension of a previous workshop paper [6] , we have enlarged our litterature review of structured peer to peer systems regarding the way they handle load imbalance to the case of distributed big data systems. We have generalized those popular approaches by proposing a core API that we have proved to be indeed also applicable to the Event Cloud middleware way of implementing a load balancing policy.

Vercors: Integrated environment for verifying and running distributed components

Participants : Ludovic Henrio, Oleksandra Kulankhina, Eric Madelaine.

It is the general prurpose of the Vercors platform to target the generation of distributed applications with safety guarantees. In Vercors, the approach starts from graphical specification formalisms allowing the architectural and behavioral description of component systems. From this point, the user can automatically verify application properties using model-checking techniques. Finally, the specified and verified component model can be translated into executable Java code. The Vercors tool suite is distributed as an Eclipse plugin. This year

• we implemented a first reliable version of the whole tool chain including generation of verifiable models and executable Java code.

• We applied the approach to several examples including Peterson’s leader election algorithm, a workflow executor, and the control and management of service composition [7] .

• A paper accepted at FASE'2016 presents an overview of this work; a research report provides the full version of the paper [20] . The theoretical background was published as a research report and an improve version is being submitted as a journal paper.

The practical implementation allowed us to improve the presentation of the theory and better evaluate it.