Enhance Resilience and QoS Awareness in Message Oriented Middleware for Mission Critical Applications

Message-oriented middleware (MOM) provides a messaging service layer between the transport and application layer of the networking protocol stack. A resilient and QoS-aware MOM system provides dependable QoS, and strives to maintain and recover a required level of messaging in the face of challenges, e.g., bursty surges in workload demand and failures in the underlay network or brokers. The QoS support and resilience mechanisms in our federated MOM system is built on several building blocks as follows: firstly resilient overlay routing supports mitigating and recovering from underlay network failure(s), secondly it is enhanced by a hybrid static/dynamic scheduling strategy which supports QoS in the network dynamics, and thirdly a novel workload allocation mechanism which minimizes the quantified risk of workload exceeding capacity of a broker, while introducing redundant mirroring of workload. This paper discusses the overall scheduling strategy and the workload allocation and mirroring mechanism we employed. Comparing with round robin maximizing resource reserve ratio, our allocation algorithm provides superior resilience in minimizing the risk of correlated workload exceeding the capacity of system.