Proactive QoS negotiation in asynchronous real-time distributed systems

Abstract

We present a fast, proactive, quality of service (QoS) negotiation algorithm called Best Effort Negotiation (or BEN), for asynchronous real-time distributed systems. BEN considers an application model where trans-node application timeliness and fault-tolerance requirements are expressed using benefit functions, and anticipated workload and system failure rates during future time intervals are expressed using adaptation functions and reliability functions, respectively. Furthermore, BEN considers an adaptation model where subtasks of application tasks are replicated at run-time for tolerating failures as well as for sharing workload increases. Given such models, the objective of the algorithm is to maximize the sum of aggregate real-time and fault-tolerance benefits during the time window of adaptation functions. Since determining the optimal solution is computationally intractable, BEN heuristically computes sub-optimal resource allocations in polynomial-time. To determine how well BEN performs, we describe another algorithm called HLC, that is inspired by the well-known Hill Climbing heuristic. We show that HLC is significantly slower than BEN. However, our experimental studies reveal that the performance of BEN, in general, is as good as that of HLC.