A Programming Language Approach
to Fault Tolerance for Fork-Join Parallelism

Mustafa Zengin and Viktor Vafeiadis

Max Planck Institute for Software Systems (MPI-SWS)

This project approaches fault tolerance of parallel computations from a programming language prespective. We consider a functional programming language with fork-join parallelism, and give it a simple high-level operational semantics that accounts for fail-stop failures, and provides means to recover from failed computations. We implement this high-level semantics with a lower level one that models processor failures and recovery more accurately.

Our Coq development defines these semantics formally and proves correspondence properties between the fault-prone semantics and standard fault-free evaluation.

More information can be found in the following paper: Sources:
Valid XHTML 1.0 Transitional