MODELING AND OPTIMIZATION OF PARALLEL AND DISTRIBUTED EMBEDDED SYSTEMS By ARSLAN MUNIR A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2012 ⃝c 2012 Arslan Munir 2 ACKNOWLEDGMENTS I would like to express my sincere gratitude to my Ph.D. advisor Dr. Ann Gordon-Ross and my Ph.D. co-advisor Dr. Sanjay Ranka for their guidance and support during the course of my Ph.D. I sincerely appreciate the considerable amount of time and effort they invested in guiding me with my research. I would also like to acknowledge Dr. Gregory Steffan for inviting me for visiting research at the University of Toronto (UofT), Ontario, Canada. Many thanks to my Ph.D. committee members Dr. Janise McNair, Dr. Greg Stitt, and Dr. Prabhat Mishra for their comments that helped in improving the quality of this dissertation. This work was supported by the Natural Sciences and Engineering Research Council of Canada (NSERC) and the National Science Foundation (NSF) (CNS-0834080, CNS-0953447, and CNS-0905308). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the NSERC and the NSF. 3 TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................. 3 LIST OF TABLES ...................................... 12 LIST OF FIGURES ..................................... 14 ABSTRACT ......................................... 18 CHAPTER 1 INTRODUCTION ................................... 20 1.1 Embedded Systems Applications ....................... 23 1.1.1 Cyber-Physical Systems ........................ 24 1.1.2 Space .................................. 24 1.1.3 Medical ................................. 25 1.1.4 Automotive ............................... 26 1.2 Embedded Systems Applications Characteristics .............. 28 1.2.1 Throughput-Intensive .......................... 29 1.2.2 Thermal-Constrained .......................... 30 1.2.3 Reliability-Constrained ......................... 30 1.2.4 Real-Time ................................ 31 1.2.5 Parallel and Distributed ......................... 31 1.3 Embedded Systems — Hardware and Software ............... 32 1.3.1 Embedded Systems Hardware .................... 33 1.3.1.1 Sensors ............................ 34 1.3.1.2 Sample-and-Hold circuits and A/D converters ....... 34 1.3.1.3 Processing units ....................... 34 1.3.1.4 Memory subsystems ..................... 35 1.3.1.5 D/A converters ........................ 36 1.3.1.6 Output devices ........................ 36 1.3.2 Embedded Systems Software ..................... 36 1.3.2.1 Operating system ...................... 36 1.3.2.2 Middleware .......................... 37 1.3.2.3 Application software ..................... 37 1.4 Modeling — An Integral Part of the Embedded System Design Flow ... 38 1.4.1 Modeling Objectives .......................... 40 1.4.2 Modeling Paradigms .......................... 42 1.4.2.1 Differential equations .................... 42 1.4.2.2 State machines ....................... 43 1.4.2.3 Dataflow ........................... 43 1.4.2.4 Discrete event-based modeling ............... 44 1.4.2.5 Stochastic models ...................... 44 1.4.2.6 Petri nets ........................... 45 4 1.4.3 Strategies for Integration of Modeling Paradigms .......... 45 1.4.3.1 Cosimulation ......................... 46 1.4.3.2 Code integration ....................... 47 1.4.3.3 Code encapsulation ..................... 47 1.4.3.4 Model encapsulation ..................... 47 1.4.3.5 Model translation ....................... 47 1.5 Dissertation Contributions ........................... 47 1.6 Relationship to Published Work ........................ 50 1.7 Dissertation Organization ........................... 51 2 OPTIMIZATION APPROACHES IN DISTRIBUTED SINGLE-CORE EMBEDDED WIRELESS SENSOR NETWORKS ........................ 53 2.1 Architecture-Level Optimizations ....................... 55 2.2 Sensor Node Component-Level Optimizations ................ 57 2.2.1 Sensing Unit .............................. 57 2.2.2 Processing Unit ............................. 58 2.2.3 Transceiver Unit ............................. 59 2.2.4 Storage Unit ............................... 59 2.2.5 Actuator Unit .............................. 60 2.2.6 Location Finding Unit .......................... 60 2.2.7 Power Unit ................................ 60 2.3 Data Link-Level Medium Access Control Optimizations ........... 61 2.3.1 Load Balancing and Throughput Optimizations ........... 61 2.3.2 Power/Energy Optimizations ...................... 63 2.4 Network-Level Data Dissemination and Routing Protocol Optimizations . 65 2.4.1 Query Dissemination Optimizations .................. 65 2.4.2 Real-Time Constrained Optimizations ................ 68 2.4.3 Network Topology Optimizations ................... 68 2.4.4 Resource Adaptive Optimizations ................... 69 2.5 Operating System-level Optimizations .................... 69 2.5.1 Event-Driven Optimizations ...................... 69 2.5.2 Dynamic Power Management ..................... 70 2.5.3 Fault-Tolerance ............................. 70 2.6 Dynamic Optimizations ............................ 71 2.6.1 Dynamic Voltage and Frequency Scaling ............... 71 2.6.2 Software-Based Dynamic Optimizations ............... 72 2.6.3 Dynamic Network Reprogramming .................. 72 3 AN APPLICATION METRICS ESTIMATION MODEL FOR DISTRIBUTED EMBEDDED WIRELESS SENSOR NETWORKS ........................ 73 3.1 Application Metrics Estimation Model ..................... 75 3.1.1 Lifetime Estimation ........................... 75 3.1.2 Throughput Estimation ......................... 81 3.1.3 Reliability Estimation .......................... 82 5 3.1.4 Models Validation ............................ 83 3.2 Experimental Results ............................. 84 3.2.1 Experimental Setup .......................... 84 3.2.2 Results ................................. 85 3.2.2.1 Lifetime ............................ 86 3.2.2.2 Throughput .......................... 87 3.2.2.3 Reliability ........................... 87 3.3 Concluding Remarks .............................. 88 4 MARKOV MODELING OF FAULT-TOLERANT DISTRIBUTED EMBEDDED WIRELESS SENSOR NETWORKS ........................ 89 4.1 Related Work .................................. 91 4.2 Fault-Tolerance Parameters .......................... 95 4.2.1 Coverage Factor ............................ 95 4.2.2 Sensor Failure Probability ....................... 96 4.2.3 Sensor Failure Rate .......................... 97 4.3 Fault-Tolerant Markov Models ......................... 97 4.3.1 Fault-Tolerant Embedded Sensor Node Model ............ 98 4.3.2 Fault-Tolerant EWSN Cluster Model ................. 101 4.3.3 Fault-Tolerant EWSN Model ...................... 104 4.4 Results ..................................... 105 4.4.1 Experimental Setup .......................... 106 4.4.2 Reliability and MTTF for an NFT and an FT sensor node ...... 107 4.4.3 Reliability and MTTF for NFT and FT EWSN clusters ........ 113 4.4.4 Reliability and MTTF for an NFT and an FT EWSN ......... 117 4.5 Concluding Remarks .............................. 121 5 AN MDP-BASED DYNAMIC OPTIMIZATION METHODOLOGY FOR DISTRIBUTED EMBEDDED WIRELESS SENSOR NETWORKS ................. 123 5.1 Related Work .................................. 126 5.2 MDP-Based Tuning Overview ......................... 130 5.2.1 MDP-Based Tuning Methodology for Embedded Wireless Sensor Networks ................................ 130 5.2.2 MDP Overview with Respect to Embedded Wireless Sensor Networks134 5.3 Application Specific Embedded Sensor Node Tuning Formulation as an MDP ....................................... 136 5.3.1 State Space ............................... 137 5.3.2 Decision Epochs and Actions ..................... 138 5.3.3 State Dynamics ............................. 138 5.3.4 Policy and Performance Criterion ................... 139 5.3.5 Reward Function ............................ 140 5.3.6 Optimality Equation ........................... 143 5.3.7 Policy Iteration Algorithm ........................ 144 5.4 Implementation Guidelines and Complexity ................. 145 6 5.4.1 Implementation Guidelines ....................... 145 5.4.2 Computational Complexity ....................... 147 5.4.3 Data Memory Analysis ......................... 147 5.5 Model Extensions ............................... 148 5.6 Numerical Results ............................... 152 5.6.1 Fixed Heuristic Policies for Performance Comparisons ....... 153 5.6.2 MDP Specifications ........................... 153 5.6.3 Results for a Security/Defense System Application ......... 158 5.6.3.1 The effects of different discount factors on the expected total discounted reward ................... 158 5.6.3.2 The effects of different state transition costs on the expected total discounted reward ................... 160 5.6.3.3 The effects of different reward function weight factors on the expected total discounted reward ........... 161 5.6.4 Results for a Health Care Application ................. 162 5.6.4.1 The effects of different discount factors on the expected total discounted reward ................... 162 5.6.4.2 The effects of different state transition costs on the expected total discounted reward ................... 163 5.6.4.3 The effects of different reward function weight factors on the expected total discounted reward ........... 164 5.6.5 Results