In this demonstration, we show our two practical fault-tolerant systems for stop failure and Byzantine failure that work under enormous and dynamic distributed systems such as Internet agent systems. The fault-tolerant systems are implemented on both JADE and our FIPA compliant agent execution platform (Erdoes). In the demonstration, we introduce ten laptop PCs connected with a 10-Mbps LAN, where eight agents are
running on each PC. On this system we show real faults and recoveries. For stop failure, we shut down any chosen PC at any time, and show that the agents running on it are consistently all recovered on stable machines. The audience can witness the consistency by checking graphically represented application messages exchanged among them. For Byzantine failure, we control any chosen agent on any PC, and show this controlled agent does not have any effect on the behavior of honest agents in any way. Eighty agents are executing sample application programs on the systems, and our methods are shown not to seriously slow down the execution. Technically, our fault-tolerance methods can overcome emerging new characteristics of distributed systems such as Internet agent systems: extremely large size, dynamism, and homogeneity (non-structure).