Hybrid Quantum and Classical Workload Management with Graph-based Scheduling

Jul 16, 2026

Upload by Eden

High Performance Computing (HPC) systems are extending from traditional compute to include quantum resources, enabling hybrid quantum-classical workflows for advanced optimization. The integration challenge is orchestration: remote QPUs add an external “two-queue problem” in parallel with the classic scheduler queue. The work introduces Fluence, a Kubernetes scheduler plugin using the Fluxion graph-based scheduler to support gang-scheduled placement. Evaluation across simulator and real QPUs reduces wasted node time, improves synchronization and worker utilization, and selects cost/queue-aware backends to cut cost by ~70x and time-to-result from hours to under a minute.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages11

Reading Time28 min

Abstract

Introduction

Quantum job model and the two-queue problem

Models of Quantum Resource Management

Remote QPU queues and the offline problem

Hybrid Quantum and Classical Workload Management with Graph-based Scheduling Vanessa Sochat∗ [sochat1@llnl.gov](sochat1@llnl.gov) Lawrence Livermore National Laboratory Livermore, California, USA Daniel Milroy [milroy1@llnl.gov](milroy1@llnl.gov)[ ](milroy1@llnl.gov)Lawrence Livermore National Laboratory Livermore, California, USA arXiv :2607 .09151v1 [ quant-ph] 10 Jul 2026 Abstract High Performance Computing (HPC) centers are expanding to encompass resources that extend beyond traditional computing. By extending resources to quantum computing, hybrid quantumclassical workflows tackle complex optimization problems that have never before been possible. However, integrating quantum processing units (QPUs) into cloud-native and scientific workload managers presents a unique orchestration challenge: remote quantum devices introduce a second, external queue—a “two-queue problem”—alongside the queue owned by the traditional scheduler. In this work we present Fluence, a Kubernetes scheduler plugin backed by the Fluxion graph-based scheduler, that enables informed, gang-scheduled placement for quantum-classical workloads and custom resources. We evaluate Fluence across three scenarios using AWS Braket simulators and real QPUs. First, under node contention, Fluence’s atomic gang placement all but eliminates the wasted nodetime that a default scheduler accrues by partially placing gangs. Second, we introduce a synchronization primitive for the two-queue problem in which a single producer submits a shared quantum task while consumers remain scheduling-gated, reducing worker idle time by roughly 5x under short device queues and by orders of magnitude when a real device queue stretched to hours. Third, costand queue-aware backend selection pins the cheapest or shortestqueue device satisfying a workload, cutting mean per-run cost by roughly 70x and time-to-result from hours to under a minute. Together, these results show that quantum-awareness can be added to a cloud-native scheduler without modifying user containers. 1 Introduction The high performance computing (HPC) center of the future isan autonomous and hydrid environment for scientific discovery, provisioning resources for scientific workloads that cross environments. A heavily accelerating technology is quantum computing, with applications moving from experimental to real world use-cases [21] . In the same way that a user can request accelerators such as Graphical Processing Unit (GPU) devices, quantum devices will be added to the resource graph in the HPC center of the future. Current quantum processors are not ready for real scientific workloads because the do not have error correction. This issue was coined the Noisy Intermediate-Scale Quantum (NISQ) regime by Preskill [17] and persits through present day. To mitigate these device limitations, research has adopted a hybrid quantum-classical frameworks that encompass a group of algorithms called Variational Quantum Algorithms (VQAs) [7] . Under this paradigm, classical resources to handle error minimization and parameter updates, and quantum ∗ Corresponding Author devices are handed discrete tasks to work on. In simple terms a classical process constructs a parameterized quantum circuit, submitsit to a Quantum Processing Unit (QPU) for execution, and receives a set of measurement outcomes for further interpretation. The circuit’s parameters can then be adjusted and the process repeated until a defined convergence is reached. Canonical examples of this approach include the Quantum Approximate Optimization Algorithm (QAOA) [11] and the Variational Quantum Eigensolver (VQE) [16], which target combinatorial optimization and ground-state energy estimation, respectively. From a workload scheduling standpoint, a quantum job becomes a classical job that has a quantum dependency. While different workflow designs can be imagined, for example sequential steps of classical and quantum, for VQAs the primary need

FAQs

What causes the “two-queue problem” in hybrid quantum-classical scheduling?

Remote quantum devices submit work into a vendor-owned second queue, while the traditional workload manager maintains its own queue. Scheduling must coordinate dependencies across both systems, creating synchronization challenges.

What is Fluence and what does it add to Kubernetes scheduling?

Fluence is a Kubernetes scheduler plugin backed by the Fluxion graph-based scheduler. It enables informed gang-scheduled placement and supports custom resources for quantum-classical workloads without changing user containers.

How does Fluence improve performance and cost when QPU queues are short or long?

It uses an atomic gang placement strategy to reduce wasted node time, introduces a synchronization primitive to lower worker idle time by roughly 5x under short device queues, and performs cost- and queue-aware backend selection to reduce mean per-run cost by about 70x and time-to-result from hours to under a minute.

Jul 16, 2026

Upload by Eden