Control and Data Plane Programming­driven SDN Management and Operation: A Top­Down Approach

Internet Service Providers (ISPs) and network operators face numerous challenges to adapt their forwarding devices in order to meet new use case requirements. Network managers and engineers often need to implement agile network management frameworks, traffic steering processes, or resource distribution techniques that try to improve the performance of their networks.

Network equipment vendors also lack technological advancements that allow new features to improve time to market. The advent of Software­Defined Networking (SDN) opened new possibilities for academia and industry to improve network optimization based on centralized path computation, end­to­end network slicing, distributed resource allocation, policy frameworks, network virtualization, and multiple other research fields within SDN.

The evolution of programmable networks pointed out some drawbacks of the first generation SDN, which triggered new data plane programming language and Data­controller plane interface (D­CPI) protocol. Control and data plane programming are two of the primary focal points of this thesis that tries to show the potential of exploiting both to improve network resource utilization, visibility, and management.

Therefore, this thesis focuses on the challenges of deploying, operating, and managing the present and Next­generation Software­Defined Networks. The research procedure followed in the thesis is described as top­down, referring to two main fields: a downward research focus across SDN architectural layers and, secondly, programming abstractions.

As described below, the first research field studied in this thesis belongs to the topmost SDN and programming abstraction layers. Ass the thesis progresses to subsequent parts, the focus will transition towards the lowest layers. The beginning of this thesis organizes the research from topics that represent the highest levels of SDN architectural layers (management and application planes) and programming abstractions.

Reviewing the literature for SDN high­level programming abstractions helps in finding plenty of obstacles to facilitate the initiation into programming control and data planes. In fact, the first part of the thesis focuses on SDN visual programming and high­level automation for modular SDN networks.

The first research area demonstrates that higher­level abstractions help deploy network programs by abstracting the control or data plane functionalities with visual models. Similarly, the additional research on modular control and data planes provides efficient management interfaces to build SDN networks from scratch.

It separates network functionalities into correlated control and data plane modules that are merged to enable an automatic network deployment. Then, the research focus continues towards investigating the possibilities of SDN­based slicing and resource distribution for railway communication. The slicing application demonstrates that it is possible to isolate different tenants and also distribute resources accordingly at runtime.

The core research of this topic focuses on application and control planes. The third part of the thesis is focused on integrating In­band Network Telemetry (INT) in Multiprotocol Label Switching (MPLS) networks and reusing telemetry information to improve the performance of service provider networks.

The telemetry framework implemented can provide new traffic steering methods based on, INT as well as new label and path latency verification techniques. The main implementations of the research conducted in this part focus on control and data planes. Finally, the last part of the thesis focuses on predicting the influence of data plane programming elements in data plane processing latency.

The results demonstrate that it is possible to utilize a reliable method to predict the latency increase in programmable data planes.