Enhanced distributed recovery blocks: a unified approach for the design of safety-critical distributed systems

A novel method is given for dealing with both hardware and software faults in a distributed system and the authors illustrate how the method copes with communication failures between the interacting distributed processors. The work has been designed using the Occam programming language and implemented on a network of transputers. This work is being extended to more complex control applications and shows very good results. The mechanism used is based on distributed recovery blocks (K.H. Kim, H.O. Welch, 1989). It is argued that distributed recovery blocks (DRB) are well suited for real-time control applications. DRB are based on the standard method of recovery blocks. The enhancements incorporated within DRB include the concurrent execution of the try blocks over a distributed network of processing nodes and the dynamic reconfiguration of nodal operations in the event of a fault. The systems proposed, takes the basic DRB and introduces extra acceptance tests to reduce the changes of Byzantine type errors and is termed an Enhanced DRB (EDRB)