Loading...

Shared Resource Management in DAG-Based Task Sets on Mixed-Criticality Multi-core Systems

Jafari, Sahar | 2025

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 58384 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Hessabi, Shaahin; Safari, Sepideh
  7. Abstract:
  8. In safety-critical systems, software tasks with varying criticality levels must execute in a coordinated manner under strict timing constraints on a multicore platform to ensure overall system safety. These tasks typically have temporal and logical dependencies and are not independent; in practice, mixed-criticality systems rely on structures of interdependent tasks with different criticality levels, which can be modeled using directed acyclic graphs (DAGs). Graph-based tasks may require access to shared resources during execution, and such access must preserve data integrity while preventing deadlocks and chained blocking. However, prior research has largely overlooked the critical issue of resource sharing for task graphs. The complexity of precedence constraints makes task synchronization analysis challenging, and the presence of a fair protocol for managing shared resources and limiting task blocking times is essential. In this study, we first propose a new model for representing mixed-criticality task graphs with shared resource constraints. We then introduce an innovative protocol that ensures fair access to shared resources and reduces task blocking times. This protocol is designed for complex task graph structures and guarantees exclusive resource management. In addition to the resource protocol, we present a method for clustering and mapping tasks onto processing cores, enabling the system to support both heavy and light graphs while minimizing interference among tasks with different criticality levels. Clustering is performed under a dynamic, criticality-aware scheduler to maintain system safety. In mixed-criticality systems, preserving the safety of high-criticality tasks is paramount, yet providing quality of service for low-criticality tasks is also important. Therefore, under critical conditions, the degraded execution of low-criticality tasks and task migration ensures a minimum quality of service, allowing these tasks to retain a relatively high level of service even under stress. Finally, task schedulability is analyzed using a demand-bound function, and path-based blocking analysis is conducted. Simulation results demonstrate that the proposed approach improves schedulability and maintains quality of service compared to prior work, confirming its high performance in ensuring safety while providing minimum service guarantees. Simulation results demonstrate that the proposed approach improves schedulability and preserves QoS compared to prior work, confirming its effectiveness in ensuring safety while delivering guaranteed levels of service
  9. Keywords:
  10. Mixed-Criticality Emdedded Systems ; Directed Acyclic Graph (DAG)-Based Applications ; Shared Resources Management ; Service Quality ; Blocking Probability ; Clustering

 Digital Object List

  • محتواي کتاب
  •   view
  •   download

 Bookmark

No TOC