Chercheur Doctorant H/F

PhD Position: Distributed and Resilient Task Placement on Heterogeneous and Dynamic Edge-Computing Infrastructures Research field: Computer Science Research Work Scientific Context Intensive applications [2] generate large volumes of data that need to be processed under real-time constraints, especially with the rise of connected objects and diverse digital streams (images, videos, texts, sounds, speech). Computation offloading requires efficient and decentralized scheduling strategies, particularly for applications such as object recognition, speech processing [8], or augmented reality [1]. Cloud computing [3] offers virtually unlimited capacity but has limitations, notably for processing sensitive information and latency constraints stemming from best-effort interconnections. To address this, edge computing [10], by being closer to data sources, enables greater responsiveness and reduces information exposure. Utilizing these heterogeneous and dynamic computing resources enhances the responsiveness required for intensive applications. This heterogeneity is reflected in the diversity of computing units (CPUs, GPUs, FPGAs, MPPAs), their specifications (instruction throughput, cache sizes), and their network accessibility. Additionally, the variability of applications and datasets increases platform complexity [5, 4], necessitating adapted and decentralized scheduling strategies. Thesis Subject This PhD project focuses on task scheduling in highly dynamic environments, particularly where the available resource pool evolves during execution. Under these conditions, meeting quality of service (QoS) and user experience (QoE) constraints depends on efficient utilization of computing infrastructures. However, task scheduling in such a context is complex due to the diversity of deployed resources (sensors, drones, robots, vehicles, cloud, etc.), each with its own availability and processing power constraints. Task planning is further complicated by fluctuating connectivity and environmental disruptions (interference, hardware failures, weather conditions). To address these challenges, we propose a distributed task scheduler capable of leveraging all available heterogeneous resources, avoiding excessive centralization in the cloud, and ensuring greater responsiveness and resilience. Several scientific challenges must be tackled: • Real-time task allocation optimization, integrating automatic or controlled duplication to hide communication times and enhance system resilience. This optimization must also consider critical application constraints, where processing cannot be interrupted. • Facilitating collaboration among heterogeneous resources without exclusively relying on a centralized cloud, utilizing a dynamic multi-layered infrastructure, and potentially integrating a private cloud depending on data sensitivity. • Scheduling robustness in a dynamic environment where connectivity is intermittent and resources fluctuate. The algorithm must adapt to connection losses and resource availability variations while ensuring processing continuity. • Intrinsic scheduler resilience, requiring fast recovery strategies, intelligent redundancy, and self-adaptation to system disturbances to guarantee efficient and robust scheduling under any constraints. Previously, we developed a task scheduler for edge-cloud infrastructures, aiming to minimize delays (QoS). The approach integrates task duplication to enhance system responsiveness. We implemented a Mixed Integer Linear Programming (MILP) model enabling task duplication based on constraints. This model was implemented using three Ant Colony Optimization (ACO) metaheuristics: Ant Colony System (ACS), Max-Min Ant System (MMAS), and Rank-Based Ant System (RBAS). Various benchmarks covering multiple use cases were defined. Initial results show quick convergence towards optimal or near-optimal solutions. Building on these preliminary works, the objective is to design a distributed scheduler capable of real-time analysis of heterogeneous resources and opti mizing task allocation, including partial or full duplication based on critical ity. The goal is to ensure responsiveness and robustness in highly dynamic environments where neither future requests nor resource availability are pre dictable. Our initial results suggest that ACO algorithms are well-suited to this problem. In particular, ACS allows for rapid convergence towards near optimal scheduling solutions [9]. Furthermore, ACO can adapt to dynamic environments as long as variations remain moderate [7]. Artificial intelligence, through multi-agent systems, represents a relevant complementary approach [6]. These agents will analyze the state of the environment and resources, as well as the decisions made by the ACO. Correlating with past situations will enable contextual adaptation and increased algorithm reactivity. Work Plan 1. Study task placement techniques for (quasi-) real-time applications on dynamic edge-computing infrastructures. 2. Efficiently and robustly orchestrate tasks on an edge infrastructure. Based on the research already conducted by the supervision team, propose a distributed and dynamic task placement model based on ACO. Using previous research, implement a decentralized monitoring system based on agents to collect resource status information, enabling fine-tuned ACO scheduler configuration. All research will result in technical outputs evaluated through simulation and real-world platform testing. Expected Scientific/Technical Output The research will lead to publications in top-tier international conferences and journals. The thesis will result in the development of a dynamic and distributed scheduler capable of orchestrating tasks on highly dynamic edge platforms. Lab presentation CESI LINEACT (UR 7527), Laboratory for Digital Innovation for Businesses and Learning to Support the Competitiveness of Territories, anticipates and accompanies the technological mutations of sectors and services related to industry and construction. The historical proximity of CESI with companies is a determining element for our research activities. It has led us to focus our efforts on applied research close to companies and in partnership with them. A human-centered approach coupled with the use of technologies, as well as territorial networking and links with training, have enabled the construction of cross-cutting research; it puts humans, their needs and their uses, at the center of its issues and addresses the technological angle through these contributions. Its research is organized according to two interdisciplinary scientific teams and several application areas. Team 1 "Learning and Innovating" mainly concerns Cognitive Sciences, Social Sciences and Management Sciences, Training Techniques and those of Innovation. The main scientific objectives are the understanding of the effects of the environment, and more particularly of situations instrumented by technical objects (platforms, prototyping workshops, immersive systems) on learning, creativity and innovation processes. Team 2 "Engineering and Digital Tools" mainly concerns Digital Sciences and Engineering. The main scientific objectives focus on modeling, simulation, optimization and data analysis of cyber physical systems. Research work also focuses on decision support tools and on the study of human-system interactions in particular through digital twins coupled with virtual or augmented environments. These two teams develop and cross their research in application areas such as • Industry 5.0, • Construction 4.0 and Sustainable City, • Digital Services. Areas supported by research…

Location: Lingolsheim, FR

Posted Date: 5/2/2025