Timestamp system for causal broadcast communication

Abstract

In unreliable asynchronous distributed systems with failures, achieving a causal view of the system across all processes is a challenging task. The Causal Reliable Broadcast (CRB) abstraction is used to solve this task. When CRB is implemented with algorithms that use logical vector clocks to timestamp broadcast events, the causal relationships between broadcast events can be detected with maximal accuracy. However, this timestamping mechanism used by CRB might not be useful for systems that need to reason about the causal relationships among both broadcast and delivery events. To address this challenge, the paper proposes a Causal Timestamp System (CTS) based on vector clocks that timestamps broadcast and delivery events capturing with maximal accuracy the causal relationships among those events. CTS simplifies the formal verification and testing of implementations of CRB algorithms based on CTS. Additionally, a new Global State Monitoring (GSM) algorithm is proposed, tailored to a distributed system that uses CRB with CTS. GSM enables finer-grained assessment of global states and application-dependent predicates of that system. We clarify these concepts with an IoT example.