Chasing the FLP Impossibility Result in a LAN or How Robust Can a Fault Tolerant Server Be?

Peter Urban, Xavier Defago and Andre Schiper: Chasing the FLP Impossibility Result in a LAN or How Robust Can a Fault Tolerant Server Be? Keywords: replication, atomic broadcast, consensus, measurements, robustness, high load, LAN, FLP impossibility Abstract: Fault tolerance can be achieved in distributed systems by replication. However, Fischer, Lynch and Paterson have proven an impossibility result about consensus in the asynchronous system model, and similar impossibility results exist for atomic broadcast and group membership. We investigate, with the aid of an experiment conducted in a LAN, whether these impossibility results set limits to the robustness of a replicated server exposed to extremely high loads. The experiment consists of client processes that send requests to a replicated server (three replicas) using an atomic broadcast primitive. It has parameters that allow us to control the load on the hosts and the network, as well as the timeout value used by our heartbeat failure detection mechanism. Our main observation is that the atomic broadcast algorithm never stops delivering messages, not even under arbitrarily high load and very small timeout values (1 ms). So, by trying to illustrate the practical impact of impossibility results, we discovered that we had implemented a very robust replicated service.