Exploiting Controllers to Adapt Message-Oriented Middleware

Publish/subscribe distributed systems rely on brokers as their central element. These brokers have many customizable parameters typically configured before execution and remain static throughout the system’s operation. However, adjusting these parameters dynamically at runtime can offer significant advantages. It enables brokers to better adapt to changing workloads, optimize resource usage, and align with evolving application requirements. Despite these benefits, dynamically customizing broker parameters poses challenges due to the number of adjustable parameters, inherent software complexity, difficulty properly setting the parameters, and the need for dedicated APIs. This paper explores using control theory to adjust brokers’ parameters dynamically. The investigation involves implementing and tuning various controllers, such as classical PID, variants of PID and unconventional controllers drawn from cloud computing and network control flow. These controllers are used to fine-tune the behaviour of a commercial broker dynamically. Ultimately, the control-theoretical broker is rigorously assessed using metrics to identify the most effective controller and tuning method. The findings of this research have practical implications, contributing to the effective management of messaging systems by demonstrating the practical application of control theory concepts and mechanisms to dynamically adapt the broker to changing environmental conditions.