Section: Overall Objectives

Overview

The CONVECS project-team addresses the rigorous design of concurrent asynchronous systems using formal methods and automated analysis. These systems comprise several activities that execute simultaneously and autonomously (i.e., without the assumption about the existence of a global clock), synchronize, and communicate to accomplish a common task. In computer science, asynchronous concurrency arises typically in hardware, software, and telecommunication systems, but also in parallel and distributed programs.

Asynchronous concurrency is becoming ubiquitous, from the micro-scale of embedded systems (asynchronous logic, networks-on-chip, GALS – Globally Asynchronous, Locally Synchronous systems, multi-core processors, etc.) to the macro-scale of grids and cloud computing. In the race for improved performance and lower power consumption, computer manufacturers are moving towards asynchrony. This increases the complexity of the design by introducing nondeterminism, thus requiring a rigorous methodology, based on formal methods assisted by analysis and verification tools.

There exist several approaches to formal verification, such as theorem proving, static analysis, and model checking, with various degrees of automation. When dealing with asynchronous systems involving complex data types, verification methods based on state space exploration (reachability analysis, model checking, equivalence checking, etc.) are today the most successful way to detect design errors that could not be found otherwise. However, these verification methods have several limitations: they are not easily accepted by industry engineers, they do not scale well while the complexity of designs is ever increasing, and they require considerable computing power (both storage capacity and execution speed). These are the challenges that CONVECS seeks to address.

To achieve significant impact in the design and analysis of concurrent asynchronous systems, several research topics must be addressed simultaneously. There is a need for user-friendly, intuitive, yet formal specification languages that will be attractive to designers and engineers. These languages should provide for both functional aspects (as needed by formal verification) and quantitative ones (to enable performance evaluation and architecture exploration). These languages and their associated tools should be smoothly integrated into large-scale design flows. Finally, verification tools should be able to exploit the parallel and distributed computing facilities that are now ubiquitous, from desktop to high-performance computers.