Access Type

Open Access Dissertation

Date of Award

January 2014

Degree Type


Degree Name



Computer Science

First Advisor

Nathan Fisher


For many hand-held computing devices (e.g., smartphones), multiple operational modes are preferred because of their flexibility. In addition to their designated purposes, some of these devices provide a platform for different types of services, which include rendering of high-quality multimedia. Upon such devices, temporal isolation among co-executing applications is very important to ensure that each application receives an acceptable level of quality-of-service. In order to provide strong guarantees on services, multimedia applications and real-time control systems maintain timing constraints in the form of deadlines for recurring tasks. A flexible real-time multi-modal system will ideally provide system designers the option to change both resource-level modes and application-level modes. Existing schedulability analysis for a real-time multi-modal system (MMS) with software/hardware modes are computationally intractable. In addition, a fast schedulability analysis is desirable in a design-space exploration that determines the "best" parameters of a multi-modal system. The thesis of this dissertation is:

"The determination of resource parameters with guaranteed schedulability for real-time systems that may change computational requirements over time is expensive in terms of runtime. However, decoupling schedulability analysis from determining the minimum processing resource parameters of a real-time multi-modal system results in pseudo-polynomial complexity for the combined goals of determining MMS schedulability and optimal resource parameters."

Effective schedulability analysis and optimized resource usages are essential for an MMS that may co-execute with other applications to reduce size and cost of an embedded system. Traditional real-time systems research has addressed the issue of schedulability under mode-changes and temporal isolation separately and independently. For instance, schedulability analysis of real-time multi-mode systems has commonly assumed that the system is executing upon a dedicated platform. On the other hand, research on temporal isolation in real-time scheduling has often assumed that the application and resource requirements of each subsystem are fixed during runtime. Only recently researchers have started to address the problem of guaranteeing hard deadlines of temporally-isolated subsystems for multi-modal systems. However, most of this research suffers two fundamental drawbacks: 1) full support for resource and application level mode-changes does not exist, and/or 2) determining schedulability for such systems has exponentialtime complexity. As a result, current literature cannot guarantee optimal resource usages for multi-modal systems. In this dissertation, we address the two fundamental drawbacks by providing a theoretical framework and associate tractable schedulability analysis for hard-real-time multi-modal subsystems. Then, by leveraging the schedulability analysis, we address the problem of optimizing a multi-modal system with respect to resource usages.

To accelerate the schedulability analysis, we develop a parallel algorithm using message passing interface (MPI) to check the invariants of the schedulable real-time MMS. This parallel algorithm significantly improves the execution time for checking the schedulability (e.g., our parallel algorithm requires only approximately 45 minutes to analyze a 16-mode system upon 8 cores, whereas the analysis takes 9 hours when executed on a single core). However, even this reduction is still expensive for techniques such as design-space exploration (DSE) that repeatedly applies schedulability analysis to determine the optimal system resource parameters. Today's massively parallel GPU platforms can be a cost-effective alternative for scaling the number of computer nodes and further reducing the computation time. An efficient GPU-based schedulability analysis can also be used online to reconfigure the system by re-evaluating schedulability if parameters change dynamically. In this dissertation, we also extend our parallel schedulability analysis algorithm for a GPU. Finally, we performed a case-study of radar-assisted cruise control system to show the usability of multi-modal system which consists of fixed priority non-preemptive tasks.