Hybrid MOEAs beat standalone algorithms for service placement in computing continuum

A new academic paper by Sergi Vivo, Carlos Guerrero, and Isaac Lera tackles the challenge of multi-objective service placement in the computing continuum—the distributed edge-fog-cloud architecture. The key innovation is not a new algorithm from scratch, but a systematic hybridization of existing MOEAs using an island model where multiple algorithms co-evolve and periodically exchange solutions. The researchers conducted two experimental campaigns: the first compared four state-of-the-art MOEAs (NSGA-II, NSGA-III, U-NSGA-III, and SMS-EMOA); the second tested a complementary hybrid configuration based on NSGA-II, MOEA/TS, and MOCPO. This hybrid approach naturally aligns with the decentralized nature of computing continuum, enabling scalable parallel execution across distributed nodes.

Across 30 independent runs, the hybrid method consistently outperformed most standalone algorithms, with statistical tests confirming significant improvements in Pareto-front quality indicators (GD, IGD, HV, S, STE). Importantly, the paper introduces a traceability-oriented analysis using genetic load to measure each island's contribution, revealing non-uniform contributions among algorithms. This provides interpretable evidence of effective hybrid cooperation. For professionals deploying AI at the edge, this research offers a tractable way to optimize service placement—reducing latency and infrastructure costs—by leveraging existing MOEAs in a collaborative, distributed fashion.