Virtual Network Slice Planning

The evolution towards 5G consists of managing highly dynamic network slices and making decisions related to the provisioning of resources in an as a service and cost-efficient fashion. The forthcoming 5G mobile system will rely essentially on cloud computing and network slicing. A network slice is defined as a chunk of virtual computing and connectivity resources configured and provisioned for a particular service, according to its characteristics and requirements. The success of cloud computing and network slicing hinges on the ongoing research activities relevant to the virtualization of mobile network functions, the efficient allocation of virtual resources (e.g. VCPU, VMDISK), and the optimal placement of Virtual Network Functions (VNFs), composing the network slices. In this vein, the aim of this dissertation, in the field of networking technology, is essentially centered on the design of a virtual network slice planning framework, allowing to efficiently place VNFs and satisfy the requirements of users and cloud service providers in terms of Quality of Service (QoS) and cost. To this end, the research questions are as follows: How to deploy efficiently the needed VNFs in order to fulfil service requests in a QoS-aware and cost-efficient manner? Is it possible to predict the behaviour of User Equipments (UEs)? How to cope with the non-uniform distribution of service requests? How to perform an integral slice planning of the EPC? The research questions are answered through experiments that entail: i) the development of a simulation tool dubbed ''Network Slice Planner", ii) the design of VNF placement algorithms, taking into consideration the predictability of service usage and the non-uniform distribution of service requests, iii) the abstraction of the Long Term Evolution (LTE) workload generation process, iv) the design of performance models of the whole LTE network, and v) the optimal placement of vEPC/5G Core virtual instances. On this basis, further research directions are identified and could be undertaken to advance the current state of knowledge on virtual network slice planning