| With the advance of network computing, networked real-time systems are increasingly important to everyday life. The development of these real-time applications with network considerations is very difficult. Techniques and tools helping engineers design, implement and test these applications, are highly desirable.; In this dissertation, we present: (1) A modeling technique: Extended Fuzzy-Timing Petri Net (EFTN) with great modeling power suitable for networked real-time systems. (2) Three Performance analysis techniques for EFTNs: (a) Model checking on transition firing sequences based on temporal logic formulas, which can be used to check the possibility that reliability and real-time QoS requirements are being satisfied in networked real-time systems. (b) Net reductions, which is used to attack the “state explosion” problem and make performance analysis efficient for large-scale networked real-time systems. (c) Simulation by using Design/CPN, which serves as test-beds and predicts the performance of networked real-time systems in design. (3) EFTN modeling and analysis techniques are applied to the following two case studies: (a) Distributed multimedia synchronization, and (b) Networked Virtual Environments. (4) Software architecture considerations for EFTNs: we present an example illustrating how to build hierarchical EFTN models.; EFTN modeling and analysis can provide great insight into networked real-time system design and dynamic behavior, estimate system performance, and thus help to improve design of systems. With EFTN as the formal modeling technique, performance analysis techniques for EFTNs, and results of applying these modeling and analysis techniques, this dissertation contributes to an integrated framework to conduct performance analysis for large-scale network-based real-time systems with high efficiency. |
