Universiti Teknologi Malaysia Institutional Repository

The deadline-based scheduling of divisible real-time workloads on multiprocessor platforms

Chuprat, Suriayati (2009) The deadline-based scheduling of divisible real-time workloads on multiprocessor platforms. PhD thesis, Universiti Teknologi Malaysia, Faculty of Science.

[img]
Preview
PDF (ABS)
56Kb
[img]
Preview
PDF (TOC)
65Kb
[img]
Preview
PDF (CHAP1)
76Kb

Abstract

Current formal models of real-time workloads were designed within the context of uniprocessor real-time systems; hence, they are often not able to accurately represent salient features of multiprocessor real-time systems. Researchers have recently attempted to overcome this shortcoming by applying workload models from Divisible Load Theory (DLT) to real-time systems. The resulting theory, referred to as Real-time Divisible Load Theory (RT-DLT), holds great promise for modeling an emergent class of massively parallel real-time workloads. However, the theory needs strong formal foundations before it can be widely used for the design and analysis of real-time systems. The goal of this thesis is to obtain such formal foundations, by generalizing and extending recent results and concepts from multiprocessor real-time scheduling theory. To achieve this, recent results from traditional multiprocessor scheduling theory were used to provide satisfactory explanations to some apparently anomalous observations that were previously made upon applying DLT to real-time systems. Further generalization of the RT-DLT model was then considered: this generalization assumes that processors become available at different instants of time. Two important problems for this model were solved: determining the minimum number of processors needed to complete a job by its deadline; and determining the earliest completion time for a job upon a given cluster of such processors. For the first problem, an optimal algorithm called MINPROCS was developed to compute the minimum number of processors that ensure each job completes by its deadline. For the second problem, a Linear Programming (LP) based solution called MIN-� was formulated to compute the earliest completion time upon given number of processors. Through formal proofs and extensive simulations both algorithms have been shown to improve the nonoptimal approximate algorithms previously used to solve these problems.

Item Type:Thesis (PhD)
Additional Information:Thesis (Ph.D (Matematik))- Universiti Teknologi Malaysia, 2009; Supervisor : Prof. Dr Shaharuddin Salleh
Uncontrolled Keywords:real-time workloads, multiprocessor real-time systems
Subjects:Q Science > QA Mathematics
Divisions:Science
ID Code:13589
Deposited By: Ms Zalinda Shuratman
Deposited On:17 Feb 2012 04:14
Last Modified:17 Feb 2012 04:24

Repository Staff Only: item control page