From Clarity to E iciency for Distributed Algorithms

This article describes a very high-level language for clear description of distributed algorithms and optimizations necessary for generating e cient implementations. The language supports high-level control ows where complex synchronization conditions can be expressed using high-level queries, especially logic quanti cations, over message history sequences. Unfortunately, the programs would be extremely ine cient, including consuming unbounded memory, if executed straightforwardly. We present new optimizations that automatically transform complex synchronization conditions into incremental updates of necessary auxiliary values as messages are sent and received. The core of the optimizations is the rst general method for e cient implementation of logic quanti cations. We have developed an operational semantics of the language, implemented a prototype of the compiler and the optimizations, and successfully used the language and implementation on a variety of important distributed algorithms.