Energy-Efficient Multivariate Privacy-Aware RF Spectrum Reservation in Wireless Virtualization for Wireless Internet of Things

Abstract

Next generation communication systems are expected to be heterogeneous and to have varying Quality of Service (QoS) requirements for different Internet of Things (IoT) applications. Wireless virtualization is regarded as an emerging technology capable of handling the ever growing resource demands by those heterogeneous IoT applications and systems by enabling wireless resource sharing through slicing. Scarce wireless resources and expensive physical infrastructure are leased to business units. The business units do not own the infrastructure, but rather gain access through a subleasing process. However, due to the lack of efficient management of wireless resources, energy efficiency of the physical infrastructure cannot be guaranteed. A viable solution to this problem is to anticipate resource needs and proactively implement energy efficient measures to ensure QoS requirements of users are met using a reservation scheme. Previous spectrum reservation approaches do not consider energy efficiency in the reservation process. In this paper, we present an adaptive resource reservation scheme for wireless network virtualization. The proposed scheme relies on an informed estimate of resource needs by a combination of aggregated event data from multiple contributors, and prediction based on previous allocation data. We present a Privacy aware Aggregation Model (PrivAgg) that relies on the truthfulness of contributors by providing secure enrollment and communication. We also present a prediction algorithm, Volume and Bandwidth-conditioned Spectrum Selective Moving Average (VBSSMA), that enforces a multivariate filter for the allocation history in order to increase the accuracy of prediction. We evaluate the performance of the reservation scheme and algorithms using theoretical analysis and numerical results from extensive simulations. Numerical results over multiple configurations for the weight of the aggregator and predictor components shows that VBSSMA results in up to 27% less allocation cost and 4% less error than existing Volume-conditioned Spectrum Selective Moving Average reservation approach.