Causality Diagrams using Hybrid Vector Clocks

Abstract: Causality in distributed systems is a concept that has long been explored and numerous approaches have been made to use causality as a way to trace distributed system execution. Traditional approaches usually used system profiling and newer approaches profiled clocks of systems to detect failures and construct timelines that caused those failures. Since the advent of logical clocks, these profiles have become more and more accurate with ways to characterize concurrency and distributions, with accurate diagrams for message passing. Vector clocks addressed the shortcomings of using traditional logical clocks, by storing information about other processes in the system as well. Hybrid vector clocks are a novel approach to this concept where clocks need not store all the process information. Rather, we store information of processes within an acceptable skew of the focused process. This gives us an efficient way of profiling with substantially reduced costs to the system. Building on this idea, we propose the idea of building causal traces using information generated from the hybrid vector clock. The hybrid vector clock would provide us with a strong sense of concurrency and distribution, and we theorize that all the information generated from the clock is sufficient to develop a causal trace for debugging. We post-process and parse the clocks generated from an execution trace to develop a swimlane on a web interface, that traces the points of failure of a distributed system. We also provide an API to reuse this concept for any generic distributed system framework.