The Consensus Problem in Unreliable Distributed Systems (A Brief Survey) t

Agreement problems involve a system of processes, some of which may be faulty. A fundamental problem of fault-tolerant distributed computing is for the reliable processes to reach a consensus. We survey the considerable literature on this problem that has developed over the past few years and give an informal overview of the major theoretical results in the area. 1. A g r e e m e n t P r o b l e m s To achieve reliability in distributed systems, protocols are needed which enable the system as a whole to continue to function despite the failure of a limited number of components. These protocols, as well as many other distributed computing problems, requires cooperation among the processes. Fundamental to such cooperation is the problem of agreeing on a piece of da ta upon which the computation depends. For example, the data managers in a distributed database system need to agree on whether to commit or abort a given transaction [20, 26]. In a replicated file system, the nodes might need to agree on where the file copies are supposed to reside [19, 30]. In a flight control system for an airplane [35], the engine control module and the flight surface control module need to agree on whether to continue or abort a landing in progress. The key point here is not what the processes are agreeing on but the fact that they must all come to the same conclusion. An obvious approach to achieving agreement is for the processes to vote and agree on the majori ty value. In the absence of faults, this works fine, but in a close election, the vote of one faulty process can swing the outcome. Since distinct reliable processes might receive conflicting votes from a faulty process, they might also reach conflicting conclusions about the outcome of the election and hence fail to reach agreement. Davies and Wakerly [2] realized this diffictflty and proposed a multistage voting scheme to overcome the problem. iThis work was supported in part by the Office of Naval Research under Contract N00014-82-K-0154, and by the National Science Foundation under Grant MCS-8116678.