AI 中文总结
研究针对分布式基础设施调度器对安全态势和量子-经典工作负载支持不足的问题，提出SQUIRO框架，基于统一调度模型和设计方法，结合安全态势强化与剩余风险优化，引入后端选择器，实验表明其能确保合规、降低成本能耗并揭示后端选择差异。
AI 中文摘要
传统分布式基础设施调度器在容量、局部性和成本优化方面支持有限，对安全态势和新兴量子-经典工作负载考虑不足。随着混合量子-经典计算的发展和后量子安全需求的影响，调度器需综合考虑异构计算资源、安全约束和量子后端特性。我们提出了SQUIRO框架，基于平台无关的统一调度模型（USM）和六步调度器设计方法（SDM），可用于Kubernetes、高性能计算（HPC）和联邦环境的具体调度器推导。该框架通过硬可行性约束实现多维安全态势强化与剩余风险优化，引入电路感知量子后端选择器。在合成Kubernetes集群上的评估表明，安全模型可确保合规工作负载完全合规，全局优化在不影响准入优先级的情况下，可将基础设施成本降低多达51%，能耗降低多达63%。额外实验还表明，在相干和队列受限条件下，电路感知后端选择与朴素错误率排名存在系统差异。
英文摘要
Distributed infrastructure schedulers traditionally optimise capacity, locality, and cost, but provide limited support for security posture and emerging quantum-classical workloads. As hybrid quantum-classical computing becomes increasingly practical and post-quantum security requirements begin to affect infrastructure deployment, schedulers must jointly reason about heterogeneous compute resources, security constraints, and quantum backend characteristics. We present SQUIRO, a framework for security-aware quantum-classical scheduling based on a platform-independent Unified Scheduling Model (USM) and a six-step Scheduler Design Methodology (SDM) that together enable the derivation of concrete schedulers for Kubernetes, high-performance computing (HPC), and federated environments. The framework combines multidimensional security posture enforcement through hard feasibility constraints with residual-risk optimisation, and introduces a circuit-aware quantum backend selector that accounts for coherence margin, calibration freshness, queue pressure, and hardware capabilities through a forward-compatible colocation hierarchy. Evaluation on synthetic Kubernetes clusters shows that the security model enforces complete compliance for regulated workloads by construction, while global optimisation reduces infrastructure cost by up to 51% and energy consumption by up to 63% compared with greedy placement in underloaded scenarios, without compromising admission priorities. Additional experiments characterise the solve-time growth of the current CP-SAT formulation and show that circuit-aware backend selection systematically diverges from naive error-rate ranking under coherence- and queue-limited conditions.
CommentsSubmitted for review to Future Generation Computer Systems
