Efficient and Reliable Communication in Distributed Embedded Systems
Doctoral thesis, 2002
The use of distributed computing elements has grown in the embedded systems arena, consequently the use of shared communication media linking these computers has garnered increasing attention. Two prominent and contemporary media sharing approaches include variations of controlled and contention based media access paradigms, with the time-triggered (TT) and the event-triggered (ET) approaches, respectively, being prominent manifestations of these paradigms. For these mentioned TT and ET approaches, the thrust of this thesis is on investigating and analyzing efficient realizations of such.
A desired attribute in embedded systems, with safety and real-time requirements, is the basic capability to coordinate and synchronize system time and events. This directly relates to the establishment of a predictable communication base, which subsequently becomes the basis to provide for predictable communication at the system level. Focusing on bus-based communication protocols, we present a novel synchronization approach targeting efficiency and low communication overhead as the main drivers for TDMA environments. Existing techniques require explicit transfer of node id information for synchronization. In this novel approach, the synchronization process utilizes the implicit information in each node's unique message length as node identifier. Furthermore, our initial startup synchronization approach is fault-tolerant, and has a bounded startup time. We also present a re-synchronization strategy that incorporates recovering nodes into synchronization.
The event-triggered and the time-triggered media access paradigms have spawned discrete followings with much debated pros and cons regarding their relative flexibility, bandwidth efficiency and predictability features. The event-triggered approach is commonly perceived as providing high flexibility. Similarly, the time-triggered approach is expected to provide a higher degree of predictable communication access to the media. One part of this thesis is to objectively and quantitatively assess the capabilities and limitations of each of these paradigms. More importantly, we quantify the spread of their differences, and provide system design guidelines for suggested best usage for each approach. The focus of this work is on response times of the communication system, and the schedulability of the communication system in collaboration with tasks in the nodes. Focusing on efficiency, the second component of this thesis, deals with introducing modifications in the time-triggered approach to efficiently accommodate event-triggered communication using the time-triggered operations as a base.