Adaptive reinforcement learning improves resource and quality balance in computing clusters
ARGOS: Reinforcement Learning-Driven Multidimensional Elasticity for Service Orchestration in the Computing Continuum
Distributed, Parallel, and Cluster ComputingArtificial Intelligence
Summary
Data-driven services running across various types of computing devices must juggle how good their results are, how many resources they use, and how much they cost. This balance is tricky when the system runs near full capacity because it has to handle changing demands without dropping quality too much. The authors introduce ARGOS, a smart controller that learns over time to adjust multiple factors like quality and resource usage together, instead of just changing resources. Tests show ARGOS adapts better than traditional methods, especially when the system is busy and resources alone can’t be scaled up.
What this means in practice
- •For cloud infrastructure teams: Manage multi-tenant data services dynamically balancing analytics quality and resource limits under changing demand with reinforcement learning.
- •For edge computing operators: Optimize service placement and scaling on heterogeneous edge devices to maintain quality when hardware capacity is constrained.
Authors
Javier Mateos-Bravo, Sergio Laso, Juan Luis Herrera, Ilir Murturi, Pantelis Frangoudis, Schahram Dustdar
Abstract
Data-intensive services in the Computing Continuum must balance analytics quality, resource usage, and cost across heterogeneous nodes with limited and uneven capacity. This balance becomes especially difficult when resource scaling reaches capacity limits, because changes in demand and cluster pressure must then be absorbed without violating client-defined quality ranges. Existing orchestrators mainly adapt resources, placements, or replicas, while analytics requirements such as coverage, sample, and freshness remain fixed. This article presents ARGOS, the Adaptive Reinforcement Learning-Driven Governance for Orchestrated Services, an end-to-end controller that formulates multidimensional elasticity as a per-request Markov decision process over analytics quality and cluster pressure, supported by capacity-aware admission. ARGOS is evaluated under controlled workloads and time-varying multi-tenant arrivals on a heterogeneous cluster. Across the controlled scenarios, the deep reinforcement learning policies consistently outperform the non-learning baselines and approach the independently tuned best-fixed reference. A separate live evaluation reports improvements over the static midpoint under realistic and saturated arrivals, with no recorded CPU or memory violations but remaining coverage violations. These results support deep reinforcement learning as an adaptive mechanism for multidimensional elasticity when resource scaling alone is insufficient.