DYNRESHPC: 4th “EuroHPC Workshop on Dynamic Resources in HPC”

Link: https://sites.google.com/uji.es/dynreshpc-workshop/home

The current static usage model of HPC systems is becoming increasingly inefficient. This is driven by the continuously growing complexity and heterogeneity of system architectures, combined with the increased usage of coupled applications, the need for strong scaling with extreme scale parallelism, and the increasing reliance on complex and dynamic workflows.

Consequently, we see a rise in research on systems supporting dynamically varying resources, middleware software, and applications, which can adjust resource usage dynamically to extract maximum efficiency. By providing intelligent global coordination of resource usage through runtime scheduling of computation, network usage, and I/O across all components of the system architecture, dynamic resource HPC systems can maximize the exploitation of the available resources while at the same time minimizing the overall makespan of applications in many, if not most, cases.

Dynamic resource management on large-scale HPC platforms is part of several EuroHPC projects. It poses highly interdisciplinary challenges concerning, among others, the applications supporting dynamic resources, extending dynamic resource provisioning to the full system stack, I/O malleable systems, new applications, and I/O scheduling, etc.

AMTE: Asynchronous Many-Task systems for Exascale 2025 (AMTE 2025)

Link: https://amte-workshop.github.io 

Supercomputers have begun to operate at exascale speed, and a tremendous amount of work has been invested in identifying and overcoming the challenges leading up to this moment. These challenges include load-balancing, fast data transfers, and efficient resource utilization. Asynchronous Many-Task (AMT) models and runtime systems have shown that it is possible to address these challenges by providing additional mechanisms such as oversubscription, task/data locality, shared memory, and data dependence-driven execution. This workshop explores the advantages of AMT programming on modern and future HPC systems. It will gather developers, users, and proponents of these models and systems to share experiences, discuss how they meet the challenges posed by today’s heterogeneous Exascale system architectures, and explore opportunities for increased performance, robustness, and full-system utilization.

WSCC: Third International Workshop on Scalable Compute Continuum (WSCC 2025)

Link: https://wscc.di.unipi.it

The recent proliferation of the Internet of Things (IoT) and edge devices, together with the nowadays consolidated Cloud Computing, is contributing to the widespread yet fragmented diffusion of heterogeneous computing resources. Moving from IoT to cloud, more powerful computational resources are available, but with increased communication latency and pricing. On the other way around, moving from cloud to IoT, more pervasive and reactive resources are available, but they are also more resource and energy-constrained and heterogeneous. The “Compute Continuum” paradigm promises to take the current state of affairs one step further: by managing the heterogeneity and dynamism of such widespread computing resources, it aims at simplifying the execution of distributed applications improving data locality, performance, availability, adaptability, energy management, and many other non-functional features. This is made possible by overcoming resource fragmentation and segregation in tiers, enabling applications (or their components) to be seamlessly executed and relocated along a continuum of resources spanning from the edge to the cloud. 

By distributing resources all around, the emerging Compute Continuum paradigm supports the execution of data-intensive applications as close as possible to data sources and end users. Besides vertical and horizontal scaling, this paradigm also offers more detailed adaptation actions that strictly depend on the specific infrastructure component (e.g., reducing power to improve battery life, migrating computation to exploit specific features such as GPUs and FPGAs). This enables the enhancement of latency-sensitive applications, the reduction of bandwidth consumption, the improvement of privacy protection, and the development of novel services aimed at improving living, health, safety, and mobility. All of this should be achievable by IoT services and application developers without having to worry about how and where the developed components will be executed. This is in contrast with the features of a paradigm where applications and the underlying infrastructure appear to be seamlessly intertwined. A profitable development of a Scalable Compute Continuum entails addressing important technological challenges, which can guarantee appropriate levels of abstraction and efficiency.

The increased complexity cannot be manually managed: efficient, scalable, and possibly autonomous mechanisms and strategies are needed to manage complexity, unleashing the true potential offered by the Compute Continuum. Further challenges exist. Distributed systems tailored for Compute Continuum can run applications targeting the needs of a large pool of entities (e.g., sensors, controllers, actuators, end users moving within a city). The inherent dynamism of the surrounding environment may negatively affect applications, e.g., by introducing unpredictable network congestion, load spikes, and resource unavailability. Proactive, autonomous, and infrastructure-aware management is desirable, if not mandatory, calling for novel interdisciplinary approaches that exploit optimization theory, control theory, machine learning and artificial intelligence methods.

In this landscape, the workshop is willing to attract contributions in the area of distributed systems with particular emphasis on support for geographically distributed platforms and autonomic features to deal with variable workloads and environmental events, and take the best of heterogeneous and distributed infrastructures.

PECS: Workshop on Performance and Energy Efficiency in Concurrent and Distributed Systems

Link: https://pecs-workshop.github.io/ 

Nowadays, concurrent and distributed systems play a role in a variety of computing applications. This is supported by the widespread adoption of computing architectures based on multi-core processors, multiprocessors or distributed computing units, which have become a standard de-facto at any system scale, from high-end servers to personal/mobile devices. However, the increasing level of hardware parallelism and heterogeneity has made concurrent and distributed systems even more complex to design, analyze and optimize. In particular, this concerns performance and energy efficiency aspects, which are known to be highly interrelated. Nonetheless, they play a key role, even because of the increasing global electricity demand by the IT sector.

In concurrent and distributed systems, energy efficiency and performance can be affected in a complex way by various factors, such as the concurrent use of computing resources, the distributed communication, the presence of (distributed) data dependencies and, ultimately, the need for synchronization of concurrent threads/processes. However, these and other factors also offer opportunities to be explored in the design of techniques and tools for improving energy efficiency and performance.

The main objective of PECS is establishing a venue for both academia and industry to discuss challenges and perspectives, and to explore methods, techniques and tools, for energy efficiency and performance analysis and optimization in concurrent and distributed systems.