| Garbage collection (GC) is becoming increasingly prevalent within the programming landscape. The use of garbage collection in real-time systems has also gained increasing attention given the major advantages in productivity and safety. Unfortunately, the progress in GC for mainstream applications and languages does not solve the problems of memory management for embedded real-time systems. The fundamental problem is that conventional GC techniques do not provide the timing and memory use predictability that are basic real-time program requirements.; This dissertation presents a systematic approach to real-time garbage collection (RTGC) for real-time systems. Given a real-time system with a bounded workload, the goal is to calibrate the control of RTGC in a systematic and automatic way. We use an approach that separately models (1) the performance of collector operations and (2) the garbage collection load offered by real-time tasks. It allows prediction of system behavior from knowledge of component behavior and environment specifications. This knowledge provides designers tools to use GC in a predictable way.; Three major topics are presented: a system model that outlines the systematic RTGC approach, a GC cost model that guarantees garbage collection activities' worst-case execution time (WCET), and GC-integrated scheduling and schedulability analysis mechanisms that compute feasible GC execution parameters satisfying real-time timing requirements.; A modified Boehm-Demers-Weiser (BDW) GC implemented in Mono 1.1.16, [96], demonstrates the use of the RTGC approach with a status router real-time application. |
