[論文レビュー] Quantum Data Center Infrastructures: A Scalable Architectural Design Perspective

Hassan Shapourian, Eneet Kaur+5 more|arXiv (Cornell University)|Jan 9, 2025

Cloud Computing and Resource Management被引用数 3

ひとこと要約

この論文は、光スイッチを用いて複数のQPUsを相互接続するスケーラブルな量子データセンター網を設計し、スイッチ中心型とサーバー中心型のトポロジー、エンタングルメント生成プロトコル、ネットワーク対応オーケストレータとシミュレーションベンチマークを導入する。

ABSTRACT

This paper presents the design of scalable quantum networks that utilize optical switches to interconnect multiple quantum processors, facilitating large-scale quantum computing. By leveraging these novel architectures, we aim to address the limitations of current quantum processors and explore the potential of quantum data centers. We provide an in-depth analysis of these architectures through the development of simulation tools and performance metrics, offering a detailed comparison of their advantages and trade-offs. We hope this work serves as a foundation for the development of efficient and resilient quantum networks, designed to meet the evolving demands of future quantum computing applications.

研究の動機と目的

• Motivate the need for scalable, networked quantum data centers to overcome the qubit count limits of individual QPUs.
• Propose scalable network architectures (switch-centric and server-centric) inspired by classical data center designs.
• Develop physical-layer models and entanglement generation protocols compatible with QDCs.
• Introduce a network-aware quantum orchestrator to translate circuit-level jobs into optical-switch configurations.
• Provide simulation benchmarks to evaluate latency and fidelity across architectures.

提案手法

• Develop physical-layer models for entanglement generation protocols (emitter-emitter, emitter-scatterer, scatterer-scatterer) with spin-photon interfaces.
• Analyze three entanglement protocols under different encodings (Fock-space, polarization, time-bin) and quantify fidelity and generation rate.
• Propose switch-centric (e.g., Clos) and server-centric (e.g., BCube) QDC network topologies and map intra-rack vs inter-rack operation to NIR and telecom wavelengths.
• Introduce a network-aware quantum orchestrator that precompiles instructions for optical switches and quantum hardware based on circuit descriptions and topology.
• Incorporate simulation and benchmarking of random quantum circuits on a Clos topology to assess end-to-end performance.]
• 研究課題（research_questions）：
• How can quantum data centers achieve on-demand, all-to-all connectivity while minimizing expensive quantum hardware?
• What network architectures (switch-centric vs server-centric) best balance scalability, fidelity, and entanglement distribution efficiency for large QDCs?
• How do different ebit generation protocols and encodings impact end-to-end entanglement rates and fidelities in practice?
• What role does a network-aware orchestrator play in translating quantum circuits into network configurations?
• How do architectural choices affect end-to-end latency and fidelity in simulated quantum workloads?

Figure 2: An abstraction for the quantum communication ports of a QPU based on the interface between the communication qubit and photonic qubit. The optical circulator is used to separate the incoming (red) photonic qubit or coherent pulse from the (blue) outgoing photonic qubit. The energy level st

実験結果

リサーチクエスチョン

• RQ1 How can quantum data centers achieve on-demand, all-to-all connectivity while minimizing expensive quantum hardware?
• RQ2 What network architectures (switch-centric vs server-centric) best balance scalability, fidelity, and entanglement distribution efficiency for large QDCs?
• RQ3 How do different ebit generation protocols and encodings impact end-to-end entanglement rates and fidelities in practice?
• RQ4 What role does a network-aware orchestrator play in translating quantum circuits into network configurations?
• RQ5 How do architectural choices affect end-to-end latency and fidelity in simulated quantum workloads?

主な発見

• The paper proposes two architectural categories—switch-centric and server-centric—based on classical data-center networking principles to enable scalable QDC interconnects.
• It analyzes three entanglement generation protocols (emitter-emitter, emitter-scatterer, scatterer-scatterer) across encodings, deriving expressions for end-to-end generation rates and fidelities under loss.
• Switch-centric layouts use non-blocking photonic interconnects with optical switches to enable near all-to-all connectivity, while server-centric designs rely on modular entanglement distribution with dedicated hardware.
• Intra-rack communications operate at near-infrared frequencies while inter-rack communications use telecom wavelengths, with non-degenerate entanglement sources and quantum frequency conversion as needed.
• A network-aware orchestrator is introduced to translate circuit-level quantum jobs and topology into precompiled control instructions for optical switches and hardware, enabling coordinated distributed quantum computation.
• Performance benchmarking on a Clos topology with random quantum circuits is conducted to evaluate average network latency and quantum fidelity.

Figure 3: Quantum data center network architectures: (a) Clos topology, (b) BCube topology, as a representative architecture for switch-centric and server-centric topologies, respectively. We use two types of switches: Telecom switches (dark blue disk-shaped) and near-infrared switches (blue rectang