TAVRUM: Feasibility-First Contract Matching for Fragmentation-Aware Resource Allocation in IoT–Fog Networks

Abstract

The rapid growth of Internet of Things (IoT) applications has increased the need for task-offloading mechanisms that can operate efficiently under heterogeneous Fog resources, dynamic workloads, and stringent latency requirements. A recurring limitation in existing allocation and matching approaches is that the choice of resource-bundle granularity is often treated as a secondary sizing step rather than as an integral part of the association decision. This paper presents Task-Adaptive Virtual Resource Unit Matching (TAVRUM), a contract-level framework in which each allocation decision explicitly couples a task, a Fog node, and a discrete Virtual Resource Unit (VRU) level. TAVRUM applies feasibility-first filtering to enforce multidimensional resource and deadline constraints before preference ranking, incorporates allocation waste into contract utility, and uses risk-aware admission together with event-triggered local rematching and hysteresis for dynamic conditions. Under the trace-calibrated experimental setting, TAVRUM reduces allocation fragmentation from 0.733 for the EDF-Feasible comparator to 0.656, an improvement of approximately 10.5%, while its task-outage rate is 33.27% compared with 32.89% for EDF-Feasible. This resource-efficiency gain is accompanied by higher mean latency (1.431 versus 1.216) and nearly unchanged P95 latency (3.247 versus 3.235). On the evaluated small instances, TAVRUM has a mean objective gap of 11.76% relative to the exact MILP reference. The results therefore characterize a measurable resource-efficiency–QoS trade-off rather than universal superiority across all metrics.