Cloud computing is new paradigm that attracts an increasing number of clients due to the advantages that it supplies, such as economic, management, deployment, security, etc. In this context, Cloud providers must possess efficient techniques to supply clients with computational resources over scalable networks. An optimal and fast provisioning algorithm is fundamental to achieve the above objective. Cloud providers need to minimise their provision cost whilst guaranteeing the requested users’ Service Level Agreements (SLA).
In this thesis, we will address the problematic of virtual network resource provisioning within Cloud backbone network infrastructure. Our research aims to optimise the allocation of virtual networks over a physical network while meeting the end users’ requirements and maximising the revenue of the Cloud provider. The problem of virtual network resource provisioning is NP-hard. To overcome the great complexity involved, we will propose several heuristic provisioning strategies and we will tackle the problem in three stages. In the first stage, we will consider a static virtual network embedding where peak demand is considered. We will propound a new scalable virtual network embedding strategy named VNE-AC based on the Ant Colony metaheuristic. In the second stage, we will build on the work carried out in the first stage by integrating a reconfiguration mechanism in order to improve the resource usage and enhance Cloud provider profitability. This will lead us to propose a new greedy Virtual Network Reconfiguration algorithm named VNR. Finally, in the third stage, we will consider an adaptive embedding strategy taking into consideration circulating traffic in order to avoid resource over-provisioning led by peak-demand allocation. We will introduce an adaptive virtual network resource allocation strategy named Adaptive-VNE to deal with the complexity and the inefficiency of resource allocation. The results obtained prove the efficiency of our proposed strategies.