Dynamic Resource Allocation and Its Applications to Multicore Real-Time Systems Robert Gifford

Format:

Book

Thesis/Dissertation

Author/Creator:

Gifford, Robert, author.

Contributor:

University of Pennsylvania. Computer and Information Science., degree granting institution.

Language:

English

Subjects (All):

0464.

0723.

0984.

Local Subjects:

0464.

0723.

0984.

Physical Description:

1 electronic resource (214 pages)

Contained In:

Dissertations Abstracts International 87-07A

Place of Publication:

Ann Arbor : ProQuest Dissertations and Theses, 2025

Language Note:

English

Summary:

Multicore architectures have become prevalent in real-time embedded and cyber-physical systems due to their ability to address increasing computational demands. However, these architectures introduce challenges related to shared resource interference, particularly in last-level caches and memory bandwidth, which adversely affect system predictability and performance. Traditional static resource allocation methods, which isolate resources to prevent interference, lack the adaptability necessary to respond to dynamic task requirements, resulting in suboptimal performance and reliability. This dissertation investigates dynamic resource allocation techniques across a variety of settings to overcome these limitations, emphasizing their effectiveness in improving schedulability, latency, and robustness in multicore real-time systems. It presents techniques that adaptively reallocate shared resources at runtime by leveraging knowledge of task execution behavior, such as distinct phases with varying resource needs. These methods demonstrate that dynamic resource allocation can effectively reduce contention and improve performance even under fluctuating workloads. The dissertation explores dynamic allocation in a variety of system contexts. It begins with systems consisting of independent tasks under soft timing constraints, where fine-grained resource reallocation can substantially improve throughput and responsiveness while ensuring safety constraints. It then extends to multi-mode systems-systems that adjust their tasksets and execution behavior based on changes in system state-where dynamic allocation is used to optimize resource usage across and during mode transitions. Further, it incorporates systems with real-time control tasks, where the ability to co-design controller implementations and resource allocations leads to enhanced safety and schedulability. Finally, it investigates systems with complex task interdependencies, modeled as directed acyclic graphs (DAGs), where resource allocation must be co-designed with the scheduling algorithm to ensure end-to-end timing guarantees. Together, these contributions demonstrate that dynamic resource allocation is a powerful and general technique for addressing the challenges of multicore real-time systems. By adapting to changing workloads and system conditions, dynamic allocation improves resource efficiency, predictability, and overall system performance across a wide spectrum of application scenarios

Notes:

Advisors: Phan, Linh Thi Xuan; Haeberlen, Andreas Committee members: Lee, Insup; Loo, Boon Thau; Lee, Benjamin C.; Parmer, Gabriel

Source: Dissertations Abstracts International, Volume: 87-07, Section: A.

Ph.D. University of Pennsylvania 2025

Vendor supplied data

Local Notes:

School code: 0175

ISBN:

9798276005829

Access Restriction:

Restricted for use by site license