Integrating performance-aware scheduling into Kubernetes-based hybrid quantum–classical workflows

Develop and integrate performance-aware or telemetry-driven decision-making into the Kubernetes-based orchestration stack—specifically Argo Workflows combined with Kueue—to enable predictive scheduling and automated backend selection across CPU, GPU, and quantum processing unit resources in hybrid quantum–classical workflows.

Background

The paper presents a Kubernetes-native framework that coordinates CPUs, GPUs, and quantum processing units for hybrid quantum–classical pipelines using Argo Workflows and Kueue. While Kueue provides fair, queue-based scheduling, the system currently lacks mechanisms to make scheduling decisions based on real-time performance telemetry.

In practice, effective orchestration for hybrid workloads requires backend selection and scheduling policies that adapt to factors such as circuit characteristics, queueing delays, device performance, and resource contention. The authors explicitly note the absence of predictive scheduling and automated backend selection based on telemetry, identifying the need to integrate performance-aware or telemetry-driven logic into the scheduling layer.