Research On Key Issues Of Lightweight Real-Time Communication

The switched Ethernet has been the main network interconnections of the distributed real-time systems. Open sources system software and low price commodity network hardware have been the main components to build the lightweight real-time communication system in a distributed real-time system. The main challenges of designing a multi-application supported lightweight real-time communication system are the follows: the first is to provide lightweight communication mechanism which can provide Gigabit per-second or higher network communication bandwidth; the second is to satisfy the requirements of soft real-time, hard real-time and non real-time communications over switched Ethernet; and the third is to be constructed with open sources system software and low price commodity hardware, and works under the real-time micro-kernel.Switched Ethernet can provide Gigabit per-second or higher bandwidth and the processing overhead of the traditional communication software has been the main bottleneck of the network communication system. The main purpose of lightweight communications is to provide lower network communication delay and higher communication throughput. Zero-copy communication mechanism is the main implementing method of the lightweight communication. In current, zero-copy receving, especially the zero-copy receving for IP fragments, is still the main challeange to the road of zero-copy communication. To satisfy the requirements of lightweight communications, a novel zero-copy communication mechanism is proposed, which integrates the admission control mechanism with the application-level fragmentation mechanism and the device-level defragmentation mechanism. Based on the zero-copy sending communication path, this zero-copy communication mechanism solves the problem of zero-copy receiving for IP fragments in a software method.Ethernet has been widely used in real-time communications for its population, high speed, simpleness and low price. While real-time communication based on switched Ethernet has introduced new challenges to researchers. Thus the traffic stream based traffic smoothing technique is proposed. This technique, when integrated with the global static network communication resources allocating and managing mechanism, can not only realize good real-time communication over switched Ethernet, but also utilize the higher communication bandwidth of the switched Ethernet.Realizing hard real-time communication over switched Ethernet, especially over multi-hop switched Ethernet, is still an open area. Thus, the dual-level traffic smoothing technique, which is based on the network calculus theory, is proposed. Firstly, the communication characters of each networking component in switched Ethernet are modeled by the network calculus theory and the restrained conditions of the bounded end-to-end latency through all networking components are achieved. Then the dual-level traffic smoothing technique, which combines the node-level traffic smoothing processing with the global-level traffic smoothing controlling, successfully solves the problem of hard real-time communication over single hop and multi-hop switched Ethernet.In distributed real-time systems, physically distributed and independent processors need a global common notion of system time to accomplish cooperating works. The system overhead of the traditional clock synchronization mechanisms which are based on the redundant exchanged messages and complicated clock synchronization algorithms, is not tolerant in real-time communication systems. LCSP (Lightweight Clock Synchronization Protocol), which integrates with the traffic smoothing technique and based on the real-time micro-kernel and lightweight real-time communication mechanism, is proposed. Specified types of messages are blocked or suppressed using the traffic smoothing technique during the clock synchronization duration and the determined network communication delays are achieved for clock synchronization messages. Several clock synchronization algorithms are also discussed to satisfy different precisions of clock synchronization. Finally, high precission clock synchronization is realized and minimized system overhead is also achieved during the clock synchronization duration.