Skip to content

Meeting 2026 06 15

Thomas Naughton edited this page Jul 2, 2026 · 3 revisions

Meeting - June 15, 2026

Upcoming meetings

Agenda

  • Seminar Series: ORNL - Development update (Amir Shehata)

Attendance

Person Institution
Amir Shehata Oak Ridge National Laboratory (ORNL)
Thomas Naughton III Oak Ridge National Laboratory (ORNL)
Ryan Landfield Oak Ridge National Laboratory (ORNL)
Brandon Neth Hewlett Packard Enterprise (HPE)
Michael Ferguson Hewlett Packard Enterprise (HPE)
Oliver Alvarado Rodriguez Hewlett Packard Enterprise (HPE)
Eun Kyung Lee IBM
Hiroshi Horii IBM Japan
Smriti Bajaj Dell Technologies
Yasuko Eckert AMD
Neal Erickson Quantinuum
Josh Moles IonQ
Jiri Schindler Independent
Doug Oucharek Oak Ridge National Laboratory (ORNL)
Jeff Heckey AWS
Kevin K Alice & Bob
Miwako Tsuji RIKEN
Sakshi Chhabra Indian Institute of Science (IISc)
Abdur Rahman Hatim Indian Institute of Science (IISc)
Martin Schulz Technical University of Munich (TUM)

Notes

Meeting Objectives

The meeting had two primary goals:

  1. Review logistics and objectives for the upcoming openQSE Day event.

  2. Provide technical updates on:

    • openQSE GitHub repositories
    • Hardware description schemas
    • The Quantum Framework (QFW)
    • Resource management interfaces
    • Integration with quantum hardware and simulators
    • Future architectural directions

1. openQSE Day Planning and Organization

Event Structure

Amir provided an overview of the upcoming openQSE Day agenda.

Opening Session

The event will begin with:

  • Welcome and organizational remarks
  • A keynote from Georgia Tourassi (Associate Lab Director - ORNL ) discussing ORNL's quantum strategy
  • A short introduction to openQSE activities and goals

Working Group Sessions

The majority of the day will be dedicated to focused working group activities.

Participants will be able to join up to two working groups across two separate sessions.

Morning Session

9:45 AM – 12:00 PM

Afternoon Session

Following lunch and a keynote from Laura, participants may join a second working group.

Report-Out Session

Each working group will summarize:

  • Progress
  • Findings
  • Proposed milestones
  • Inter-group dependencies

at the conclusion of the day.


Working Groups

The following working groups were announced:

Working Group Lead Organizations
Software Architecture ORNL, AMD
System Architecture & Workflows AWS
Compiler Infrastructure HPE and collaborators
Quantum Resource Management IBM

Oracle representatives will participate across all working groups.


Working Group Goals

Amir emphasized two primary objectives:

Define Annual Milestones

Each group should identify:

  • Deliverables
  • Milestones
  • Priorities

for the next year.

Define Interactions Between Groups

Since the groups are interdependent, they should also identify:

  • Inputs required from other groups
  • Outputs they provide to other groups
  • Cross-group interfaces and dependencies

to avoid isolated development efforts.


2. Repository Updates

Hardware Data Representation Efforts

Thomas Naughton presented ongoing work related to vendor-neutral hardware description formats.

Two repositories were highlighted:

qhw-data

Provides:

  • Hardware schema definitions
  • Validation mechanisms
  • Portable hardware descriptions

The goal is to establish a common, vendor-neutral representation of quantum hardware characteristics.


qhw-iqm

A concrete implementation of the schema for IQM systems.

This repository demonstrates:

  • Hardware description generation
  • Integration with IQM platforms
  • Use of the common schema

Both repositories are distributed as installable Python packages to encourage reuse and integration.


Design Philosophy

The work aims to:

  • Normalize hardware descriptions across vendors.
  • Represent topology information consistently.
  • Provide graph-based hardware models.
  • Support calibration and execution metadata.

Thomas compared the effort conceptually to HWLOC, which provides standardized hardware topology descriptions in HPC systems.


3. Slurm Developer Cluster

Thomas also demonstrated a second project based on a containerized Slurm environment.

Purpose

The project provides:

  • A local virtual HPC cluster
  • Developer-friendly deployment
  • CI/CD testing environment
  • Quantum resource experimentation

using Docker containers.


Features

The environment includes:

  • Slurm controller
  • Virtual compute nodes
  • Database backend
  • QFW integration
  • Quantum resource management interfaces

Users can quickly:

  • Launch clusters
  • Submit jobs
  • Develop and test workflows
  • Evaluate scheduling and resource management approaches

without needing access to production HPC systems.


4. Quantum Framework (QFW) Overview

Amir provided a detailed overview of the Quantum Framework architecture.

Architecture Overview

QFW consists of two major layers:

Distributed Execution Framework (DEFW)

Implemented in C and Python.

Provides:

  • Communication infrastructure
  • Process coordination
  • Service discovery
  • Remote procedure calls (RPC)

A Python interpreter is embedded within a C runtime environment.


Quantum Services Layer

Provides:

  • Hardware abstraction
  • Device access
  • Simulator access
  • Vendor-neutral execution APIs

Applications interact with services rather than directly with vendor-specific APIs.


5. Service-Oriented Design

Amir demonstrated how service APIs are defined.

Service API Model

A service API defines functions such as:

  • Synchronous execution
  • Asynchronous execution
  • Completion queues
  • Event notifications

Applications call these APIs without needing to know which hardware vendor is providing execution.


Vendor-Specific Implementations

Several backend implementations currently exist, including:

  • IQM
  • Quantum Brilliance
  • NWQSim
  • TNQVM
  • Other simulators

Each backend implements the same common API.


6. QFW Execution Model

Amir explained how QFW integrates with Slurm.

Resource Allocation

Users can allocate:

Classical Resources

Traditional HPC nodes.

Quantum Resources

Physical quantum systems or simulator clusters.

Applications may use either independently or together.


Runtime Initialization

During startup:

  1. Framework initializes.
  2. Services are launched.
  3. Service directory registers available resources.
  4. Applications discover available services.
  5. Workloads are submitted through common APIs.

This enables transparent execution regardless of backend implementation.


7. Hardware Characterization Demonstration

Amir demonstrated execution against the local IQM system.

The demonstration included:

  • Running characterization circuits
  • Collecting execution results
  • Capturing hardware timing information
  • Normalizing vendor-specific output into a common schema

The resulting data includes detailed timelines showing:

  • Job creation
  • Submission
  • Validation
  • Execution stages

This normalization framework is intended to support multiple vendors in the future.


8. Future Development Directions

Several roadmap items were discussed.

Runtime Layer

AMD and ORNL are exploring a higher-level runtime layer that would:

  • Decouple execution from backend implementations
  • Support multiple software frameworks
  • Provide more flexible execution workflows

Further discussion is planned during the July 24 workshop.


Resource Management

IBM, ORNL, MQSC, and others are working on:

  • QRMI integration
  • QDMI integration
  • Unified resource interfaces

The goal is to support multiple resource managers and execution environments.


Scheduler Integration

Future plans include:

  • Slurm support
  • Flux integration
  • PBS Pro support
  • LSF support

where feasible.


Admission Control

Planned capabilities include:

  • Rate-based policies
  • Credit-based policies
  • Resource quotas
  • Usage governance

to prevent oversubscription of quantum devices.


Device Scheduling

Future scheduling research may include:

  • Round-robin scheduling
  • Shortest-job-first
  • Longest-job-first
  • Multi-user fairness policies

for quantum hardware access.


Multi-Hardware Execution

The framework currently supports individual backends.

Future work will focus on:

  • Simultaneous multi-device execution
  • Hybrid hardware workflows
  • Multi-vendor orchestration

within a common framework.


Technical Discussion

RPC Technology Choices

Question (Jiri Schindler)

Jiri asked about the rationale behind:

  • Current RPC mechanisms

  • Python service definitions

  • Alternatives such as:

    • Protocol Buffers
    • gRPC
    • Language-neutral interface definitions

Discussion

Amir explained that the current implementation evolved historically and uses:

  • Python service definitions
  • JSON serialization
  • Internal RPC mechanisms

Jiri noted that gRPC could provide:

  • Language neutrality
  • Auto-generated client/server stubs
  • Stronger type safety
  • Better interoperability

Amir agreed that future evolution toward such technologies would be reasonable and welcomed community contributions.


Vendor-Neutral Device Access

Question (Smriti Bajaj)

Smriti asked how vendor neutrality is maintained.

Response

Amir explained:

  • Applications interact only with common APIs.
  • Vendor-specific services implement those APIs.
  • Backends translate common calls into vendor-specific operations.

This architecture allows the same application to target different quantum systems without modification.


Authentication Challenges

Several participants discussed authentication.

Current State

Currently:

  • IQM access relies on API keys.
  • Keys are configured manually.

Future Direction

The group discussed:

  • Integrating authentication with Slurm identities.
  • Mapping user identities to vendor credentials.
  • Hiding API keys from applications.
  • Creating pluggable authentication mechanisms.

Jiri emphasized that vendor participation will be necessary to develop robust authentication solutions.


Cloud vs HPC Scheduling Models

Question (Michael Ferguson)

Michael asked how cloud-style execution models fit into QFW.

Specifically:

  • Cloud providers often queue jobs internally.
  • Users submit work asynchronously.
  • Resource allocation differs from HPC scheduler models.

Response

Amir explained:

  • Current development targets Slurm-based deployments.
  • Cloud-style queueing would require additional work.
  • The framework architecture should be flexible enough to support such workflows in the future.

The topic remains an area for future exploration.


Key Outcomes

  1. openQSE Day structure and working group organization were finalized.
  2. Community members were encouraged to complete working group surveys.
  3. Hardware schema and topology repositories continue to mature.
  4. Containerized Slurm developer environments are available for experimentation.
  5. QFW now supports execution against local IQM hardware.
  6. Vendor-neutral service APIs remain a central architectural goal.
  7. Authentication and cloud scheduling remain active design challenges.
  8. Community participation through pull requests, testing, and design feedback was strongly encouraged.

Action Items

Action Item Owner
Complete working group participation survey All Members
Review and contribute to hardware schema repositories Community
Test the Slurm developer cluster environment Community
Provide feedback on QFW architecture and APIs Community
Continue QRMI/QDMI integration work IBM, ORNL, MQSC
Present runtime architecture updates during July 24 workshop AMD & ORNL
Explore authentication and identity integration approaches Vendors & Resource Management WG
Evaluate alternative RPC technologies and service definitions Community

Closing

Amir concluded by encouraging community members to:

  • Clone the repositories
  • Experiment with the tools
  • Submit pull requests
  • Offer architectural suggestions

The meeting ended with a reminder that all components remain community-driven and open for contribution.

Files

Clone this wiki locally