Cross-Layer Design Approaches Accounting for Optical Physical, Network, and Application Layers

With the rapid growth of Internet traffic, the current network architecture is facing various challenges on network resource control and management. First, emerging applications such as Video on Demand and web conferencing require higher transmission bandwidth. Second, with the development of cloud services such as self-provisioned IT services and elastic computing, the user traffic becomes more dynamic, which requires flexibility on network resource management. In addition, multiple types of resources must be made available in the cloud infrastructure to the applications in order to achieve desired Quality of Service (QoS). To deal with these challenges, efficient network optimization solutions are needed to tackle the increasing bandwidth demand and adapt to the dynamic nature of the future traffic.

Software-defined Networking (SDN) is enabling networks to be programmable by decoupling the control plane and the data plane, which are currently integrated in most network equipment. This control framework has been widely accepted as an efficient network technology capable of applying cross-layer orchestration. SDN can control and manage network resources in a more efficient way by allowing the information access and exchange between different layers. For example, application can have the knowledge of the network resource status.

In this dissertation, a cross-layer design approach is firstly presented to apply the optimization between network layer and physical layer. Routing and wavelength assignment strategies are jointly considered with the characteristics and power control of the physical layer devices in order to increase the network throughput and the signal robustness. The cross-layer optimization is also investigated between network layer and application layer. Multiple online resource allocation strategies are proposed while taking into account the requirements of the cloud applications to increase the resource utilization and reduce the link bandwidth over-provisioning.