Resumen de Self-managed resources in network virtualisation environments
Rashid Mijumbi
Network virtualisation is a promising technique for dealing with the resistance of the Internet to architectural changes, enabling a novel business model in which infrastructure management is decoupled from service provision. It allows infrastructure providers (InPs) who own substrate networks (SNs) to lease chunks of them out to service providers who then create virtual networks (VNs), which can then be re-leased out or used to provide services to end-users. However, the different VNs should be initialised, in which case virtual links and nodes must be mapped to substrate nodes and paths respectively. One of the challenges in the initialisation of VNs is the requirement of an efficient sharing of SN resources. Since the profitability of InPs depends on how many VNs are able to be allocated simultaneously onto the SN, the success of network virtualisation will depend, in part, on how efficiently VNs utilise physical network resources. This thesis contributes to efficient resource sharing in network virtualisation by dividing the problem into three sub-problems: (1) mapping virtual nodes and links to substrate nodes and paths i.e. virtual network embedding (VNE), (2) dynamic managing of the resources allocated to VNs throughout their lifetime (DRA), and (3) provisioning of backup resources to ensure survivability of the VNs. The constrained VNE problem is NP-Hard. As a result, to simplify the solution, many existing approaches propose heuristics that make assumptions (e.g. a SN with infinite resources), some of which would not apply in practical environments. This thesis proposes an improvement in VNE by proposing a one-shot VNE algorithm which is based on column generation (CG). The CG approach starts by solving a restricted version of the problem, and thereafter refines it to obtain a final solution. The objective of a one-shot mapping is to achieve better resource utilisation, while using CG significantly enhances the solution time complexity. In addition current approaches are static in the sense that after the VNE stage, the resources allocated are not altered for the entire lifetime of the VN. The few proposals that do allow for adjustments in original mappings allocate a fixed amount of node and link resources to VNs throughout their life time. Since network load varies with time due to changing user demands, allocating a fixed amount of resources based on peak load could lead to an inefficient utilisation of overall SN resources, whereby, during periods when some virtual nodes and/or links are lightly loaded, SN resources are still reserved for them, while possibly rejecting new VN requests. The second contribution of this thesis are a set of proposals that ensure that SN resources are efficiently utilised, while at the same making sure that the QoS requirements of VNs are met. For this purpose, we propose self-management algorithms in which the SN uses time-difference machine learning techniques to make autonomous decisions with respect to resource allocation. Finally, while some scientific research has already studied multi-domain VNE, the available approaches to survivable VNs have focused on the single InP environment. Since in the more practical situation a network virtualisation environment will involve multiple InPs, and because an extension of network survivability approaches from the single to multi domain environments is not trivial, this thesis proposes a distributed and dynamic approach to survivability in VNs. This is achieved by using a multi-agent-system that uses a multi-attribute negotiation protocol and a dynamic pricing model forming InPs coalitions supporting SNs resource backups. The ultimate objective is to ensure that virtual network operators maximise profitability by minimising penalties resulting from QoS violations.
