Multiple Resource Allocation for Multiprocessor Distributed Real-Time Systems

this report are those of the author(s) and should not be interpreted as representing the official policies, either expressed or implied, of the Defense Advanced Research Projects Agency, ONR, the U.S. Government or Honeywell. Computer facilities were provided in part by NSF grant CCR-8811954. Workshop on Parallel and Dist Real-Time Syst, IEEE IPPS'93 April 1993 in the queue and tries to schedule it. If the task is successfully scheduled, it is removed from the queue and executed. Otherwise, the task is unlocked so that other processors may examine it in the same fashion. This scheme is convenient for a shared memory architecture, but becomes too expensive and cumbersome when such facility is not available in hardware (for example, in distributed systems). The goal of this paper is twofold: first, to study the problem of allocating multiple resources to a single task; and second, to provide a heuristic allocation policy for selecting a resource among multiple instances of identical resources. We consider a distributed system in which each site is a uniform memory access (UMA) multiprocessor, but the scheme is easily extensible to non-UMA models. For the first problem, a three-tiered resource allocation scheme is proposed. A hierarchy of the resources is defined to reduce the probability of allocation failure as the allocation progresses. To reduce the complexity of multiple resource allocation, we group the resources into schedulable sets of resources. Since the objective in real-time systems is to meet deadlines, this approach is acceptable so long as it meets a justifiable minimum level of resource utilization. It is shown that our approach allows modularity, flexibility, and potential concurrency in scheduling. Another contribution of this paper is in explicitly ide..