In distributed hard-real-time systems, when a hardware failure occurs, the task instance in current period is usually more urgent than the subsequent ones. According to this, a novel strategy of delay in non-urgent period (referred to as DNUP) is proposed. DNUP strategy can postpone the execution of non-urgent instance as late as possible and reserve the slack time for the instance with low priority. Thus it has a better chance to complete its execution in an urgent period. Extensive simulations reveal that DNUP can improve the schedulability of periodic tasks and achieve a remarkable saving on the number of processors required with respect to several well-known fault-tolerant scheduling algorithms.