AHPC³: Accelerated HPC in the Cloud-Edge Continuum
About this session
Today we are witnessing an increasing demand for high-performance computing infrastructures. Modern applications imply the need to process computationally intensive workloads. Differently from the past, where HPC infrastructures were used by just a handful of application domains, nowadays they are requested by a plethora of domains and applications. This is mainly due to the high availability of a large amount of data. Traditionally, HPC infrastructures were sharply distinct from cloud infrastructures by the unique software and hardware requirements and by their on-premises nature.
However, in recent times such distinction is getting more and more blurry, driven by the proliferation of applications such as those pertaining to Big Data and AI/ML. Modern cloud infrastructures are getting closer to HPC systems in terms of performance capabilities and hardware specifications.
The aim of this workshop is to explore the intersection of high-performance computing and modern Cloud-Edge continuum architectures. The focus will be on how HPC can be achieved by relying on Cloud-Edge architectures. The workshop will investigate how technologies that are typically exploited in the context of cloud and edge environments, including serverless computing, microservices, and load balancers, must be adapted, tailored and managed to achieve efficient and scalable solutions able to support the execution of HPC applications.
The key topics include lightweight virtualization, dynamic execution environments, and advanced scheduling technologies that are crucial to the deployment of high-performance workflows in cloud environments, but not limited to them. In addition, the workshop will focus on orchestration and deployment techniques.
The workshop aims to attract submissions on innovative programming paradigms for high-performance Cloud-Edge computing, including network communication, data management, and strategies for fault tolerance, reliability, and security. The workshop seeks for insights on managing data-intensive workloads, heterogeneous resource management tools, and HPC applications monitoring in cloud-edge environments. Emphasis is placed on sustainability, promoting energy-efficient execution, green computing practices, and intelligent resource management to reduce environmental impact.
In addition, contributions on the potential for FPGA/GPU acceleration architectures for data flow processing and joint resource sharing mechanisms in hybrid HPC environments are welcome. Similarly, we aim at exploring the latest innovations, challenges and applications in cloud/edge continuity and hybrid cloud HPC, with a focus on accelerated environments such as FPGA and GPU architectures.
Topics of interest include, but are not limited to:
Accelerated computing architectures for HPC in the Cloud-Edge continuum
including GPUs, FPGAs, AI/ML accelerators, and heterogeneous computing platforms.
Cloud-Edge architectures & resource federation for HPC
Hybrid HPC environments and federated resource sharing.
Cloud-native technologies & execution models for HPC
Serverless computing, microservices, and task offloading for HPC.
Lightweight virtualization and execution environments
WebAssembly, unikernels, containers, and microVMs.
Orchestration, deployment, scheduling & resource allocation
for HPC workflows across the Cloud-Edge continuum.
Programming paradigms, models, & runtime systems
for High-Performance Cloud-Edge computing.
Communication, data management, stream processing, & data-intensice workloads
in Cloud-Edge environments.
Heterogeneous resource management, monitoring, and observability
for HPC Cloud-Edge applications.
Fault tolerance, reliability, security, & resilience
across the Cloud-Edge continuum.
Sustainable & energy-efficient HPC
Across Cloud, Edge, and HPC infrastructures.
Federated & distributed AI/ML
Including federated learning in the Cloud-Edge continuum.
AI/ML-driven techniques for resource management, scheduling, and orchestration
Page limit
Full papers: up to 8 pages.
Double-blind
Remove author names, affiliations, and identifying acknowledgements from the manuscript prior to submission.
Proceedings
Accepted papers are published through IEEE CPS and submitted for indexing in Scopus and other major services.
Author registration
At least one author of each accepted paper must register with a regular registration and present the paper at the conference.
Paper templates
All submissions must follow the IEEE Conference Proceedings Format (double-column, 10pt size fonts). LaTeX and Word templates for the IEEE conference format.
Download templates →These follow the same rules as the main PDP 2027 call — see the full submission guidelines for complete details.
Submit to AHPC³
Use the main EasyChair system and select this session's track.