Study On Message Architecture Of Distributed Real-Time Embedded System

Embedded system has been applied to a good many fields. Many devices, ranging from mobile phone to switch, from micro-wave oven to automatic propellant equipment of rocket etc., have an embedded control system. With the development of network technology this kind of system is heading for distributed system. However, it is rather difficult to build a distributed embedded system now, because there is not a fairly good and clear system model to direct the building process. Besides this, the usual real-time requirement of this kind of system is satisfied by real-time scheduling strategy based on global clock (GC). The method of GC is not only complicated, but limited in applicability. It is merely suitable for time-based distributed real-time system. To the promising distributed real-time embedded system (DRES), the two problems mentioned above have to be solved. They are the very problems being studied in this paper, one is how to build an exact and suitable DRES model so as to provide pretty enough guidance to system building, the other is how to meet DRES's real-time requirements not using GC, to avoid the complicated clock synchronization problem. The main work this paper did and the main results it obtained could be summarized as following: One, the structure characteristics of DRES was analyzed, and according to the key role message plays in this kind of system, the message architecture model was set up. This model is DRES's exact system model, which describes both the system's static structure and dynamic run-time characteristics. Therefore it is capable of directing the system building, verifying the design correctness of system and analyzing the performance of system beforehand etc. Two, according to the guidance of message architecture, a system building method based on message architecture was put forward. Using this method, the building of DRES may boil down to the building of relatively abstract message architecture and a series of independent distributed tasks.So it is a more reasonable building method, simplifying the system building process, reducing the building difficulities, hence the correctness and reliability of system are improved. Three, considering the fact that though each distributed node in a distributed system has no exactly same absolute clock, they have the completely same relative time, the time-accounting measure was proposed to guarantee real-time. Firstly, this measure tackles the difficult clock synchronization problem, so it is easy to realize. Secondly, its applicable area is wider than GC's. It could be applied to either time-based or event-based distributed real-time system. Besides these, it is flexible in accepting and satisfying users'alterable real-time requirements. Four, the concept of message's dynamic priority was introduced; and an integrated message scheduling strategy was adopted. This strategy is guided by message's static and dynamic priority altogether, which fullfils the real-time requirements much better. Finally, the simulation design of an example of DRES: "driving control and monitoring system of car"reveals the effectiveness and merits of message architecture model and time-accounting measure.