Novel SDN and NFV techniques for optical network orchestration

Abstract

Multilayer IP-over-WDM networks are the foundation of today's metro and core networks, and the only enabling technology for the bandwidth explosion in the 5G mobile backhaul. The planning and operation of these networks have been traditionally based on separate IP and optical approaches, largely relying on capacity over-provisioning and manual operation processes. This methodology is no longer valid for the stringent requirements in terms of bandwidth, end-to-end latency, dynamicity, scale and operation complexity of 5G networks and beyond. On the contrary, a mix of evolution and revolution is required in IP-over-WDM planning and operation principles to meet these challenges. This Ph.D. thesis contributes to this field with Novel SDN and NEFV Techniques for Optical Network Orchestration. We focus on network planning and operation of access, metro, and backbone multilayer networks, including case studies where traditional network resources are allocated along IT resources to guarantee good QoS. We also explore the importance of network automation, efficient orchestration, and agile operation as a basis for sustainable network growth. Particularly, this work covers a four-year research period on several relevant topics on multilayer optical network orchestration: (i) Network Planning, (ii) Multilayer Dynamic Operation Using Software-Defined Networking, and (iii) Dynamic VNF Placement and Operation on 5G Metro-Access Networks. First, we analyze Network Planning, discussing how multilayer networks can be planned efficiently to withstand the effect of network failures. With the emergence of the aforementioned challenges, network planning goes beyond "adding more capacity". Contributions in this part have the objective to properly plan and dimension networks that guarantee not only survivability but also end-to-end latency and QoS. Both transport and metropolitan area networks are covered in our analyses, tailoring our solutions to the specific requirements and characteristics of each network segment. Second, we explore Multilayer Dynamic Operation Using Software-Defined Networking. This new paradigm is a game-changer in terms of network management and operation. Network orchestration is studied using two different approaches in the form of PoC, leveraging Net2Plan extensions. First, the orchestration of OpenFlow-based networks is achieved by creating an off-platform application on top of the OpenDaylight controller. Then, we experiment with dynamic multilayer network operation using an active stateful multilayer PCE, based on protocols such as BGP-LS and PCEP, integrated into an ABNO architecture. Lastly, we discuss Dynamic VNF Placement and Operation on 5G Metro-Access Networks Dynamic scenarios with high mobility of users and rapid changes in service expectations (e.g., football matches, concerts, massive parades) require a special and careful operation, especially taking into account the requirements expected by 5G. We present the idea of a joint orchestration of radio and cloud resources, exploring the benefits of efficient and dynamic VNF deployment with the goal of satisfying QoS/QoE requirements. Also, a novel VNF-handover mechanism is presented, showing the benefits of proactive VNF allocation to avoid service disruption. Summarizing, this Ph.D. thesis opens the door to the so-called SDNFV. On the one hand, Part I and Part II demonstrate the capabilities to create in-operation network planning solutions to achieve the maximum performance of networks during their lifetime, enabled by proper network dimensioning and tight integration between planning tools and SDN controllers, as reported throughout this document. On the other hand, NFV presents a new framework for operating a 5G-enabled access, where computing capabilities close to the customer edge are a must, and the demonstrated NFV handover is prominent. As a next step, putting together these two paradigms will enable end-to-end service orchestration, maximizing QoS/QoE whereas optimizing costs and resource utilization.