Network slicing with prioritization mechanism using Markov chain model under user demand uncertainty

Abstract

Wireless networks are increasingly being integrated with cloud computing platforms and benefit from new paradigms. Two paradigms in this regard are software defined networking and network function virtualization. These paradigms alongside cloud integration defines the architecture of future wireless networks. In addition, it is important that future wireless networks which incorporate the software defined networking and network function virtualization provide subscribers with enhanced quality of service. In this case, an enhanced quality of service implies high throughput and ultra-low latency alongside reliable communications. These quality-of-service attributes are the network service provider and subscriber expectations from future communication networks. In meeting the target performance expectations, future wireless networks are expected to have access to sufficient bandwidth resources. The provisioning of bandwidth resources arises in the abstraction of network slices. In the logical network system, the network slice is realized aboard a shared physical network with computing resources. The network slice is synonymous to the bandwidth and can be conceived as being the basic bandwidth unit that can be allocated to a given user context i.e., the subscriber utilizing an application. The future wireless network comprises multiple slices and aims to enable the utilization of slices in a manner that meets subscriber quality of service requirements. The quality of service is an important concern that involves two aspects. The first aspect is that of ensuring that subscribers achieve the ideal throughput (in bits per second) and lowest latency (in seconds) from the network. The second aspect is related to the realization of call admission control. The aspect of call admission control relates to how subscribers initiating data calls or voice calls access network slices. This challenge is important for a network service provider deploying a network that incorporates the paradigms of software defined networking, network function virtualization alongside network slices. The research being presented proposes the incorporation of a buffer to improve the call admission control in relation to a multi-subscriber multi-priority call capable network that incorporates network slices. The performance evaluation is done through the approach of the Continuous Time Markov Chain and considers different network scenarios and contexts. The call admission control is investigated for the viii considered scenarios and contexts (seven scenarios) via the formulated metrics of call blocking probability and call dropping probability. These metrics are formulated from a state transition model comprising three call classes in a multi-subscriber scenario. In addition, the performance evaluation is executed via MATLAB numerical investigation. Performance analysis and evaluation of benefits shows that the incorporation of the buffer reduces the call blocking probability and call dropping probability by an average of (70.2–83.7) % and (51.6–83.9) %, respectively.