Section: New Results

A Chemical Approach for Autonomous Service Computing

Participants : Héctor Fernández, Marko Obrovac, Cédric Tedeschi.

Chemical Computing for the Simulation of Agile-Based Software Engineering

Participants : Héctor Fernández, Cédric Tedeschi.

In the framework of Héctor Fernández' internship at Vrije University, we applied the chemical programming model to simulate the behavior of a team developing software with Agile methods. Although an unexpected application, it has been the occasion to widen the range of applications and users of the software prototype developed during Héctor's thesis.

Scalable Atomic Capture of Molecules

Participants : Marko Obrovac, Cédric Tedeschi.

Capturing the reactants involved in a reaction constitutes one of the main challenges in the execution of chemical programs. Doing it at large scale is one of the essential barriers hindering the actual execution of chemical programs at large scale. While the problem resembles the classic resource allocation problem, it differs from it by different aspects. One of the main difference stands in the fact that the probability of a conflict varies during the course of execution. When the number of possible reactions is high, then there is no need for a complex conflict resolution scheme, as it would lead to a useless additional cost. In contrary, when this number drops, the probability of a conflict increases, and a systematic conflict resolution is mandatory to ensure at least one reaction will take place.

An adaptive protocol has been proposed, based on the dynamic combination of several strategies. Based on simulations, we have shown that, by dynamically switching from one strategy to another one, even by locally deciding which protocol to use, it is possible to combine the good properties of the strategies without suffering from their drawbacks [18] .

The work was recently extended to take several rules into account. Rules have been defined to be able, not only to choose a strategy, but also to choose the rule to be executed, with the constant objective of maximizing the number of reactions executed in a given time.

DHT-based Runtime for the Chemical Programming Model

Participants : Marko Obrovac, Cédric Tedeschi.

The development of a distributed chemical machine entered its experimental phase with the development of a software prototype containing the following building blocks:

• A distributed hash table structures the network and allows any node to communicate with any other node in a logarithmic number of hops in this logical overlay.

• On top of the distributed hash table, a set of discovery mechanisms allows to find molecules needed in reactions, whatever their location is. These mechanisms are based on complex distribution and retrieval scheme borrowed from the P2P literature.

• The atomic capture protocol described before has been fully integrated in this framework.

• The discovery of molecules has been extended in order to detect the termination of the program and to be able to send the results of the computation back to the requester.

This software prototype has been deployed over the Grid'5000 platform [36] .