Vincent Lannurien
d822cdaa6a
6.6 KiB
title, venue, year, date, tags, paper_count, draft
| title | venue | year | date | tags | paper_count | draft |
|---|---|---|---|---|---|---|
| FGCS 2024 Digest | FGCS | 2024 | 2024-01-01 | 12 | false |
12 papers selected.
Quantum-centric supercomputing for materials science: A perspective on challenges and future directions
Yuri Alexeev, Maximilian Amsler, Marco Antonio Barroca, Sanzio Bassini et al.
TL;DR — A comprehensive roadmap from IBM, national labs, and universities identifying key algorithmic, software, and hardware challenges for using quantum processors alongside classical HPC to advance materials science simulations.
Why notable — Essential reading for any researcher planning quantum-classical hybrid workflows, covering the full stack from error mitigation to application mapping at scale.
Integrating quantum computing resources into scientific HPC ecosystems
Thomas L. Beck, Alessandro Baroni 0003, Ryan S. Bennink, Gilles Buchs et al.
TL;DR — Describes the architecture and middleware decisions made at Oak Ridge National Laboratory to expose quantum devices as first-class resources within an existing HPC facility.
Why notable — One of the first concrete descriptions of a production-scale quantum-HPC integration, providing a template other facilities can follow.
Lotaru: Locally predicting workflow task runtimes for resource management on heterogeneous infrastructures
Jonathan Bader, Fabian Lehmann, Lauritz Thamsen, Ulf Leser et al.
TL;DR — Lotaru learns lightweight per-workflow runtime prediction models locally on each node using micro-benchmarks, eliminating the need for a centralized profiling service on heterogeneous clusters.
Why notable — Addresses a core bottleneck in scientific workflow scheduling with a practical, evaluated approach that works without historical traces.
The globus compute dataset: An open function-as-a-service dataset from the edge to the cloud
André Bauer 0001, Haochen Pan, Ryan Chard, Yadu N. Babuji et al.
TL;DR — Releases a large real-world dataset of function invocations across edge, campus, and cloud resources collected from the Globus Compute FaaS platform, along with workload analysis.
Why notable — Provides the community with a rare, richly annotated dataset for benchmarking distributed FaaS schedulers and studying edge-to-cloud task patterns at scale.
A survey on checkpointing strategies: Should we always checkpoint à la Young/Daly?
Leonardo Bautista-Gomez, Anne Benoit, Sheng Di, Thomas Hérault et al.
TL;DR — Surveys decades of checkpointing research and rigorously examines when the classic Young/Daly formula is optimal versus when multi-level, coordinated, or application-aware strategies outperform it.
Why notable — A definitive reference for HPC fault tolerance that unifies scattered results and provides clear guidance on choosing a checkpointing strategy for modern exascale workloads.
Scalable I/O aggregation for asynchronous multi-level checkpointing
Mikaila J. Gossman, Bogdan Nicolae, Jon C. Calhoun
TL;DR — Proposes an aggregation layer that pipelines writes across multiple memory and storage tiers asynchronously, reducing checkpoint overhead for large-scale MPI applications.
Why notable — Delivers measurable improvements in checkpoint throughput on realistic HPC applications, directly addressing the I/O bottleneck at exascale.
StructMesh: A storage framework for serverless computing continuum
Diana Carrizales-Espinoza, Dante D. Sánchez-Gallegos, José Luis González Compeán, Jesús Carretero 0001
TL;DR — Introduces a hierarchical storage abstraction that unifies data management across edge, fog, and cloud tiers for serverless workflows, supporting structured data access patterns.
Why notable — Offers a practical, evaluated solution to the data management gap in cloud-edge serverless architectures, relevant to scientific and industrial workflow deployment.
Paving the way to hybrid quantum-classical scientific workflows
Sandeep Suresh Cranganore, Vincenzo De Maio, Ivona Brandic, Ewa Deelman
TL;DR — Defines a taxonomy and reference architecture for hybrid quantum-classical workflows, mapping quantum circuit execution onto existing scientific workflow management system abstractions.
Why notable — Provides the conceptual foundations needed to extend tools like Pegasus or Swift to orchestrate quantum subroutines within larger scientific pipelines.
Online learning and continuous model upgrading with data streams through the Kafka-ML framework
Alejandro Carnero, Cristian Martín 0002, Gwanggil Jeon, Manuel Díaz
TL;DR — Extends Kafka-ML to support incremental online learning directly from streaming data topics, enabling continuous model updates without retraining from scratch in edge-cloud deployments.
Why notable — Demonstrates a full open-source framework that bridges stream processing and ML model lifecycle management, with relevance to IoT and real-time analytics pipelines.
GRAAFE: GRaph Anomaly Anticipation Framework for Exascale HPC systems
Martin Molan, Mohsen Seyedkazemi Ardebili, Junaid Ahmed Khan, Francesco Beneventi et al.
TL;DR — Uses graph neural networks trained on node telemetry to predict imminent failures in exascale HPC clusters before they occur, enabling proactive maintenance and job migration.
Why notable — Shows that temporal graph models over system topology substantially outperform per-node anomaly detection, with validation on a real pre-exascale machine.
QFaaS: A Serverless Function-as-a-Service framework for Quantum computing
Hoa T. Nguyen, Muhammad Usman 0009, Rajkumar Buyya
TL;DR — Proposes QFaaS, a broker-based FaaS platform that abstracts heterogeneous quantum hardware providers behind a unified serverless interface with automatic circuit compilation and resource selection.
Why notable — Addresses the pressing need for a cloud-agnostic quantum execution layer, laying groundwork for portable quantum applications across IBM, IonQ, and similar backends.
Enabling federated learning across the computing continuum: Systems, challenges and future directions
Cèdric Prigent, Alexandru Costan, Gabriel Antoniu, Loïc Cudennec
TL;DR — Systematically surveys the technical barriers to training federated learning models that span IoT devices, edge servers, and cloud data centers, and proposes a reference architecture addressing heterogeneity and mobility.
Why notable — A timely synthesis that clarifies open problems at the intersection of federated learning and the compute continuum, useful as a roadmap for system builders.