Planning and optimization of multilayer optical networks

Abstract

The success and ubiquity of the Internet are revealing new fundamental issues which cannot be satisfactorily addressed by the current architecture, where exponentially growing demands cannot be fulfilled by increasing network capacity over static network configurations. Driven by new applications like Internet of Things (IoT) or live/on-demand media streaming, networks have evolved into converged infrastructures accommodating services for all forms of communication. As such, the Internet provides not only the platform to move data between devices, but also requires application-/traffic-awareness and cost-efficient planning and operation processes in order to address different challenges faced by network operators. For example, topics like real-time network reconfiguration triggered by any kind of event (i.e., due to failures or scheduled maintenance) and/or historical data (i.e., traffic trends), or simplified provisioning of services for reduced time-to-market, are of paramount importance for operators. This Ph.D. thesis contributes in the field of Planning and Optimization of Multilayer Optical Networks. Multilayer optical networks have been studied since 1990s as the enabling solution to keep pace with growing Internet traffic, hence supporting ultra-high bitrate traffic aggregation and routing in the highest segments of the hierarchy. Today, they are operated based on peak-traffic capacity dimensioning and static configuration, with limited or zero dynamicity, thus under-utilizing resources. To some extent, more attention should be put in the software side in order to optimize hardware utilization. Specifically, this work comprises a five-year period of research on several relevant topics of multilayer optical networking: (i) On the Role of Open-Source in Multilayer Networking, (ii) Network Survivability, (iii) Challenges for Dynamic Operation, and (iv) Smart Inverse Multiplexing in Optical Networks. The expected outcome is to provide a set of solutions that can help operators to improve the efficiency of their networks according to their needs, e.g., throughput, quality of service (QoS), or capacity/traffic forecasting. First, a discussion On the Role of Open-Source in Multilayer Networking is presented. As demonstrated by initiatives around the software-defined networking (SDN) paradigm, open-source solutions containing no vendor-specific subjects and/or proprietary implementations foster innovation in the networking world. However, SDN is a concept related to network operation and automation. Therefore, the main contribution of this part, and by extension of this thesis, is the Net2Plan framework. Net2Plan is an open-source planning and simulation tool devoted to multilayer networks. Based on an abstract vendor-neutral network model, Net2Plan can be applied to a broad range of technologies. Moreover, network orchestration is explored through some experimental proofs-of-concept (PoCs) leveraging Net2Plan and its extensions enabling communication with OpenDaylight and Application-Based Network Operations (ABNO) controllers. Second, Network Survivability is analyzed, discussing how multilayer networks can be operated in an efficient way to minimize the effect of networks failures. Today, network protection is implemented through dedicated protection in the optical layer and over-provisioning in the IP layer. Proposals in this part aim to take advantage of reconfigurability to maximize traffic survivability with minimum resource usage and disruption times. Third, a couple of relevant Challenges for Dynamic Operation of multilayer networks are addressed, namely fairness and hardware versatility, for scenarios where the network is continuously allocating services. The former refers to the ability of the network to accommodate service requests in the network without any special penalty because of their capacity request or path length. The latter is referred to the trade-offs between node switching capabilities and service blocking. The basic idea is to increase the revenue of a given network infrastructure through efficient resource usage. Finally, the benefits of Smart Inverse Multiplexing in Optical Networks, thanks to the introduction of lightpath bundling (LB) and anycast switching (AS) as a variant of classical port trunking and load balancing mechanisms, are analyzed as a way to improve the performance of multilayer networks where legacy transponders are grouped into logical entities, thus taking advantage of the statistical multiplexing. Besides, the side effects of load balancing on applications, especially due to packet reordering, are analyzed and mitigated through a complete set of benchmarks and scheduling algorithms.