Replication and allocation of task modules in distributed real-time systems

This paper addresses the problem of replicating and allocating periodic task modules to processing nodes (PNs) in distributed real-time systems subject to task precedence and timing constraints. The probability that all tasks can be completed before their deadlines-termed as the probability of no dynamic failure (P/sub ND/)-is used as the performance-related reliability measure. Modules which are critical in meeting task deadlines are then selected using the critical path analysis. To provide the timing correctness embedded in P/sub ND/, both original and replicated task modules are not only assigned to PNs, but also scheduled on each PN so as to meet the deadlines of all tasks. The module allocation scheme uses (1) the branch-and-bound method to implicitly enumerate all possible allocations while effectively pruning unnecessary search paths; and (2) the module scheduling scheme to schedule the modules assigned to each PN. Several numerical examples are presented to illustrate the proposed scheme.<>