Off-campus UMass Amherst users: To download dissertations, please use the following link to log into our proxy server with your UMass Amherst user name and password.
Non-UMass Amherst users, please click the view more button below to purchase a copy of this dissertation from Proquest.
(Some titles may also be available free of charge in our Open Access Dissertation Collection, so please check there first.)
Real-time systems: Well-timed scheduling and scheduling with precedence constraints
Abstract
This dissertation attempts to bridge the gap between recent theoretical results from Scheduling Theory, Queueing Theory, and Operations Research and actual requirements of current and future real-time systems. The first part develops a novel approach for the timely scheduling of dynamically arriving tasks, and it is, primarily, designed for on-line schedulers of non-deterministic complex real-time systems that perform in temporary or permanent overloads. In these systems, to predict whether a task will complete by its deadline, an on-line schedulability analysis has to be carried out. The quality of the analysis, as well as its computational overheads, depends on when the analysis is performed and how many tasks are involved. Our approach to schedulability analysis, called Well-Timed Scheduling, is based on analytically derived control parameters. This approach presents a framework for on-line real-time schedulers, and it lends itself to use with different scheduling policies. Well-Timed Scheduling provides a methodical approach to quantifiable guarantees of timing constraints with potentially low scheduling overhead and high system performance. Using this approach, the ready-to-execute tasks are scheduled at an "opportune" time, rather than at arrival time or at dispatch time as in the traditional approaches. The analytical derivation of the "opportune" time is based on recent theoretical results, and it is validated through simulation. Aside from run-time benefits (e.g., low scheduling overheads), Well-Timed Scheduling is useful as a design tool. It can, for example, be used to determine the number of processors needed to achieve the required level of system's guaranteed performance for a given M/G/c real-time system. In the second part of the dissertation, we develop off-line preprocessing algorithms that enable effective and efficient on-line scheduling of task groups with different contributing values, timing constraints, resource constraints, and arbitrary precedence constraints. These algorithms derive new value densities that reflect how valuable the individual tasks and their successors are. By utilizing these reflective value densities, on-line schedulers are not required to examine the successors of ready-to-execute tasks at run-time to select the best task to schedule next. This approach greatly reduces the computational complexity of on-line schedulers, and it extends the applicability range of existing on-line scheduling algorithms for independent tasks to scheduling of task groups with arbitrary precedence constraints. Due to the separation of value density preprocessing and deadline preprocessing, the developed algorithms are equally applicable to real-time and non real-time systems. The overall goal of this dissertation is the development of efficient and effective scheduling methods for on-line scheduling of complex real-time systems in very demanding non-determininstic environments, where the best-effort algorithms are used to maximize the total accrued system value.
Subject Area
Computer science
Recommended Citation
Zlokapa, Goran, "Real-time systems: Well-timed scheduling and scheduling with precedence constraints" (1993). Doctoral Dissertations Available from Proquest. AAI9316731.
https://scholarworks.umass.edu/dissertations/AAI9316731