FNB: Fast Non-Blocking Coordinated Checkpointing Protocol for Distributed Systems

This paper presents a Fast Non-Blocking coordinated checkpointing protocol for distributed systems with the aim of minimizing the number of requests and mutable checkpoints while reducing the checkpointing latency. Our protocol relies on two mechanisms; the first one is piggybacking dependency information on computation and reply message, thereby, tracking direct, transitive and hidden dependencies among processes. The second one is popular processes; due to the communication between processes, it is more desirable that the checkpointing procedure is initiated by popular processes having more dependency information. In fact, this way may reduce the checkpointing latency and the likelihood of checkpointing halting caused by fault occurrence. We also present a simulation study that compares our protocol to CSNB protocol (Cao and Singhal Non-Blocking) and CSB.protocol (Cao and Singhal Blocking).