Deterministic Message Flow Scheduling in Industrial Internet: Message Middleware Use Case

This article mainly focuses on message flow scheduling for industrial message middleware in the integration of deterministic network with industrial internet, which is to find the optimal matching between telecom operator and message requests from industrial users. However, current deterministic network is limited to two or three layers, and the whole transmission reliability and latency involving industrial networking and service layers cannot be guaranteed, resulting in uncontrollable message scheduling reliability, deterministic latency, low channel resource utilization etc. To this end, an integer linear programming (ILP) model is formulated for message flow scheduling issue, with multi-objectives optimization of resource utilization and profit under customized message sending latencies, and deterministic reliability. A heuristic algorithm is also proposed. Furthermore, we present a Deep Reinforcement Learning based Message Flow Scheduling algorithm (DRL-MFS), which not only solves complex message flow scheduling problems, but also achieves the self-optimization automatically. Proposed message flow scheduling approaches are simulated based on data from real cloud messaging service platforms. Simulations results prove that proposed DRL-MFS is superior to heuristic, and is close to ILP model in multi-objective optimization under three urgency levels of message sending. The reliability of three methods is proven to be deterministic.