Shared-memory vs. message-passing in an asynchronous distributed environment

This paper presents two basic building blocks in an asynchronous environment, one in the shared-memory model and one in the message-passing model, that help to translate algorithms between the two models. These building blocks are equivalent in the following sense: given an algorithm in one model that utilises its appropriate building block, we are able to construct an algorithm in the other model that utilizes the equivalent building block. We demonstrate this idea by two applications. First, we solve the processor renaming problems in the shared-memory model based on the solutions of in the message-passing model. The solution is deterministic wait-free and employs atomic read and write registers. Second, we provide a randomized algorithm for flipping a biased global coin in the message-passing model based on an algorithm in the shared-memory model. Using this coin, we achieve the 'fastest' randomized algorithm solving the consensus problem against the strongest adversary with one half resilience.