Making Fault-Sensitive Algorithms Tolerate Link Failures

We present a unified approach for converting a class of distributed algorithms that are fault-sensitive into algorithms tolerating undetectable link failures in asynchronous distributed systems. Single and multiple link failures not leading to a network partition are considered. Our technique is applied on several examples and fault-tolerant algorithms are derived. On the message complexity measure, the derived fault-tolerant distributed algorithms compare well with the fault-sensitive algorithms.