Doctorat en Mathématique & Informatique

Permanent URI for this collection

https://dspace.univ-chlef.dz/handle/123456789/347

Département de Mathématique & Informatique

Université Hassiba Benbouali de Chlef,
Faculté des sciences BP 151
Hay Salem, route nationale N° 19
02000 Chlef, Algérie
Tel: 027-72-70-17
Email: Bio_uhbc@univ-chlef.dz

Browse

Browsing Doctorat en Mathématique & Informatique by Subject "dynamic adaptation"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    DYNAMIC ADAPTATION APPROACH FOR DISTRIBUTED SYSTEM BEHAVIOR
    (Mohammed Amin TAHRAOUI / Ahmed HARBOUCHE, 2026) BENDIAF, MOHAMMED LOTFI
    As distributed systems (DS) evolve, managing dynamic workloads and resource availability becomes essential for maintaining optimal performance. This thesis presents a novel Dynamic Adaptation (DA) framework designed to enhance the e”ciency and responsiveness of DS, particularly in heterogeneous computing environments. The central problem addressed in this work concerns how the performance of distributed systems can be e!ectively enhanced through dynamic adaptation, particularly at the level of a single-node multiprocessor system considered as a fine-grained unit within a distributed architecture. Furthermore, what key strategies and algorithmic approaches are essential to optimize task allocation, ensure load balancing, and minimize execution time in such heterogeneous environments. A key contribution of this work is the introduction of two innovative algorithms for dynamic task scheduling. The first, DyTAg (Dynamic Task Allocation using Dynamic Programming), focuses on optimizing task allocation in heterogeneous multiprocessor systems with independent tasks. It leverages dynamic programming to minimize makespan and balance workloads e!ectively, laying the groundwork for more advanced scheduling approaches. Building on DyTAg, the thesis introduces the Knapsack-based Algorithm Co-Scheduling Task Allocation (KaCoSTA), which integrates dynamic programming with knapsack optimization techniques to address task precedence and resource constraints. KaCoSTA dynamically adapts to system states, task priorities, and processing capabilities to maximize resource utilization (RU) and minimize makespan. Extensive experiments demonstrate its superiority over existing methods, such as Min-Min, Max-Min, and HEFT (Heterogeneous Earliest Finish Time), in both static and dynamic scenarios. Results reveal significant advancements in system adaptability, load balancing, and overall e”ciency under fluctuating workloads. By combining foundational research with practical innovation, this work provides a comprehensive solution for optimizing DS behavior in real-world applications.