Resource allocation and management under priorities based on the squatting-kicking model for multi-slice 5G networks

Abstract

The upcoming Beyond Fifth Generation networks aim to meet network services characterized by low latency and high reliability among others in different slices to provide a high-quality user experience. However, existing best-effort networking schemes that implement traditional methods of controlling and allocating network and computing resources do not meet such strict service requirements well. In International Telecommunication Union-Telecommunication sector, future services are defined as Network 2030 Services under a chartered Focus Group on Networks 2030 (FG-NET2030). The results from the FG-NET2030 analysis suggests that current networks cannot accommodate real-time applications with low latency and high bandwidth requirements. Moreover, current networks lack the capabilities to dynamically aggregate and share network resources through multiple flows, which is essential for future services. However, to satisfy the strict requirements of those services, intelligent algorithms and techniques that incorporate 5G enablers are needed to introduce novel network management systems. These intelligent algorithms shall not only result in efficient utilization of network resources but also guarantee the required quality of service for the priority slices. Moreover, cognizant of the strict latency requirements of the different services, such algorithms should include delay constraints of requests. Despite the advantages expected from future services are real-time applications, should benefit from reduced physical and logical paths between end-users and data or service hosts. However, all the above requirements are not intended for the network slicing paradigm alone. Therefore, in addition to network slicing, we want to leverage technologies and components that have features such as network programming, dynamic network reconfiguration and orchestration to enable improved performance and efficient resource management. Such technologies include NFV and SDN among others. Consequently, the main objective of this PhD thesis is to develop a service deployment algorithm that uses Squatting and Kicking techniques intelligence to effectively allocate, manage, and control slice resources under several constraints in a real-time multi-slice scenario, such as priority, bandwidth, and E2E delay with targeting to maximize the total resource usage in the substrate network. The proposed online algorithm allocates the available resource to different priority demands from source node to destination node along the routed path according to more realistic constraints, such as links' bandwidth and E2E delay. Moreover, the benefits of the new proposed algorithm will be reflected in creating real-time demands for 5G applications that are sensitive to delay, in addition to solving the resource allocation problem for large scale networks, using fewer resources and generating lower costs. Further, the proposed algorithm is adaptable to meet various QoS requirements of services, guaranteeing high QoS levels and providing high admission for higher priority classes under congested scenarios. In terms of managing bandwidth resources in a multi-slice scenario, Bandwidth Allocation Models offer improved metrics over best effort models. The proposed algorithm outperforms the others by far especially, during congested scenarios. To this end, this thesis proposes a resource allocation model called Squatting and Kicking model (SKM) to maximize the number of successfully embedded demands while maximizing the utilization in the multi-slice networks by choosing less congested paths through the efficient allocation of demands on the network. Moreover, this thesis analyzes the impact of delay constraint on the performance of an online resource allocation algorithm based on an intelligent efficient SKM, proved in this work to be the most effective up to the present time yet.