Virtual Fault Tolerance

摘要

Brian Randell had many early ideas on the way that virtual systems could make programming more fault tolerant and allow computing systems to be more dependable. He developed these ideas with an architecture for acceptance testing and backup to previous safe points on failure of a test. The architecture provided for multiple versions of methods to pass a test, and it extended to multithreaded programs and atomic transactions. We will review the principles of these systems and then comment on how they apply in today's systems. We suggest that capability architecture, an important offshoot of virtual systems, could usefully be combined with Randell's architecture. The combination could offer a means to reduce the modern anxiety over Trojan Horse attacks against hardware and software.