Hybrid Quantum-Classical Resource Scheduling: A Proposed Architecture for Next-Generation Hyperscale Data Centers

Abstract

Resource allocation in hyperscale data centers has been inflated by the complexity introduced by diversity in workloads, the ramping-up of infrastructure, and the escalating performance requisites that have transcended traditional optimization limits. This research is interested in hybrid quantum-classical paradigm architectures for resource planning by allowing for each of the optimization quantum advantages while keeping practicality in consideration. The main challenge addressed by the architecture is the fact that virtually any classical-based scheduling algorithms suffer from exponential time complexity in NP-hard allocation problems, thus affecting significant compromises in resource exploitation. The proposed methodology entails building a decomposition framework to dissect scheduling problems into quantum-solvable optimization subproblems and classical tasks of task coordination, with a quantum annealer aiding in combinatorial optimization while the classical part handles dynamic constraints and real-time execution. The architecture advocated also seeks to enhance the usefulness of quantum processing units as specialized coprocessors integrated into standard data center management stacks, making them adopt a piecemeal fashion rather than necessitating a major overhaul of the existing infrastructure. The proposed framework integrates novel encoding techniques that map data center scheduling variables to qubits, hybrid solvers, blending quantum approximate optimization with classical refinements, and adaptive problem partitioning to optimize the classical computing capacity for tackling essential subproblems on classical systems. For instance, simulation across representative hyperscale workloads delivers 34% improvement in resource utilization, 47% reduction from interval fitivities in cumbersome scheduling latencies, and 28% reduction in energy consumption vis-à-vis the best-performing classic solutions. This work lays the foundation for the research of quantum-enhanced infrastructure management alongside novel architectural patterns facilitating quantum computing's expansion into production-level data centers.