The rapid growth of connected smart devices and IoT platforms has brought safety and security concerns for emerging parallel and distributed systems to the forefront. These highly complex systems operate in a nondeterministic environment with unpredictable behaviour. If left unchecked, any fault in the system, whether due to an internal error or one imposed by the external environment, can lead to an unpredictable failure with potentially disastrous consequences. Therefore, preventing faults from turning into failures is and will be a crucial consideration for ensuring system resilience.

Behavioural types provide a strong theoretical foundation for distributed programming based on message passing, which is well-suited to describe the above systems and has been extensively used to enforce safety and security properties. However, existing approaches provide limited fault models that cannot accommodate the issues encountered in emerging real-world systems due to their size, heterogeneity, dynamic nature, and in many cases an adversarial setting. Recent applications of behavioural types, including resilience through failure recovery, are promising. However, the current state of the art of behavioural types and programming languages lacks a comprehensive account of the above-mentioned demands.

We note that static analysis will not be sufficient to reach our goal in practice. Recent progress in the field of runtime verification is very promising — in particular across host languages — and will complement our efforts to form a two-pronged approach. One of our goals, therefore, is to improve our understanding of how far static analysis can be pushed while remaining practical and what a suitable interface to runtime verification could be.

This Dagstuhl Seminar aims to bring together experts from academia and industry to enrich behavioural types such that they can be used to enhance resiliency. This involves programming languages outfitted with behavioural types to tackle the resilience of today’s emerging distributed and heterogeneous computing platforms. It will focus on recovering from failures, using quantitative measures to minimise losses due to unpredictable failure behaviour, and analysing behaviour across heterogeneous system boundaries. The format of the seminar will focus more on discussion in small groups than on single speaker presentations.

Creative Commons BY 4.0

Farzaneh Derakhshan, Robbert Krebbers, Roland Kuhn, and Nobuko Yoshida