Consensus algorithms in Software Defined Networks (SDN) for multiple controllers
Αλγόριθμοι Συναίνεσης σε Δίκτυα Ορισμένων Λογισμικών (SDN) για πολλαπλούς ελεγκτές

Doctoral Thesis
Author
Λάλου, Σταυρούλα
Date
2025View/ Open
Keywords
Δίκτυα Ορισμένα από Λογισμικό (SDN) ; Πολλαπλοί ελεγκτές ; Αλγόριθμος συναίνεσης ; Ανοχή σε σφάλματα ; Συντονισμός ; Εξισορρόπηση φόρτου ; Τοποθέτηση ελεγκτών ; ΚαθυστέρησηAbstract
This doctoral dissertation investigates the application of consensus algorithms to enhance
collaboration among multiple controllers in Software Defined Networks (SDN). The research
focuses on addressing critical challenges in SDN, such as scalability, fault tolerance, single point
of failure, and network state synchronization, by leveraging consensus protocols. Traditional
single-controller architectures face limitations in managing large-scale networks, including
bottlenecks, overhead, and vulnerability to failures. In contrast, multi-controller architectures offer
improved flexibility, scalability, and reliability. However, effective coordination and
synchronization among distributed controllers remain essential for optimal network performance.
The study conducts a comparative analysis of existing consensus protocols and mechanisms,
evaluating their performance under identical simulation environments. A novel consensus
mechanism based on the Raft algorithm is proposed, designed to ensure stable, consistent, and
efficient network state synchronization across controllers. The mechanism supports high
throughput, dynamic view changes, fault tolerance, and controller synchronization, demonstrating
superior performance compared to existing alternatives. Additionally, the research explores load
balancing techniques to optimize resource utilization in the control plane, addressing challenges
such as latency and packet loss during load migration.
Furthermore, the dissertation addresses the Controller Placement Problem (CPP), proposing
a heuristic greedy algorithm to minimize end-to-end latency and reduce maximum latency between
controllers and switches. This approach aims to enhance network performance by strategically
positioning controllers and mitigating queuing delays. The study emphasizes the practical
implementation of these solutions, providing a comprehensive framework for improving
scalability, reliability, and efficiency in SDN architectures.
By integrating consensus algorithms, load balancing techniques, and optimized controller
placement strategies, this research contributes to advancing SDN management, offering robust
solutions for modern network challenges. The findings and proposed mechanism are
experimentally validated, demonstrating their effectiveness in achieving stable and high-
performing SDN environments.