Yonga: An Adaptive Metaheuristic-Based Microservice Placement Approach for Low-Resource Distributed Systems

A microservice architecture enables the decomposition of applications into loosely coupled, independently deployable components, each responsible for a specific function or service. This provides for distribution of services across heterogeneous domains with diverse resource capacities and interconnection characteristics. In low resource settings, characterised by limited compute power, unreliable and inconsistent network connectivity, high communication costs, and frequent power outages, microservices can offer distributed workload processing and mitigate the effects of failures through environment-aware placement strategies. This research paper presents Yonga, an adaptive metaheuristic-based microservice placement approach for low-resource distributed systems. Yonga balances optimality and adaptability by employing metaheuristics to explore the search space and react to dynamic environmental conditions where deterministic approaches may fail. It leverages available compute resources and periodic network state information to formulate and constrain a multi-objective optimization problem. To solve this problem, Yonga employs NSGA-II to generate a set of Pareto-optimal solutions and applies weighted scalarization to determine the final placement. Experimental results show that Yonga achieves up to 9.3× higher throughput in heterogeneous deployments, remains within 1.2× of Binpack while outperforming other heuristics in a homogeneous CloudLab environment, and sustains up to 7× performance gains under degraded network conditions.