Optimum Controller Placements in SD-WAN Deployment Case of Ethio Telecom
No Thumbnail Available
Date
2021-10
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
The explosive increase of data traffic with the proliferation of devices leads to massive interconnections.
As a result, the difficulty of network management operations and configurations,
service provisioning, and heterogeneous networks due to continuous network deployments to
meet the demand of ubiquitous connections evolved the network programmability concept as a
solution.The SDN decoupling of the control plane from the data plane with logically centralized
and programmable network came with opportunities and also challenges such as scalability and
reliability that addressed with Multiple controller systems to improve network performance.
The number of controllers needed for a given network topology and where it should be placed
optimally might vary with different use cases. For these varieties of use cases, different formulations
of mathematical models with different optimization models are addressed in many
research works. This thesis work focused on optimum controller placement for the WAN network
topology of Ethio telecom optimizing propagation latency and controller imbalance with
Constraint of controller capacity and the metrics trade-off analysis.
The optimum controller placement to minimize controller number by the constraints of propagation
delay, controller capacity, traffic estimates, and load balance. Performance evaluation
based on these parameters based optimization model of the heuristic approaches like Simulated
Annealing and K-Medoid is performed. In this work, the trial and error way of determining
controller number taking as the input parameter is replaced with determining it from controller
capacity and node weights (flow requests from nodes). Also introduced the concept of load
balance based on Jain’s fairness index to measure how load is fairly distributed.
At last, the impact of controller capacity assessment on parameters of the optimum placement
investigation showed controller capacity affects the placement metrics. As a result, K-Medoid
showed improvement of 32% to 90% in load balance taking the same latency as a reference and
12.5% to 29% node to controller latency taking the same controller imbalance as a reference.
Finally, optimum controller placement was identified and shown on google earth.
Description
Keywords
Software-Defined Networking, SD-WAN, OpenFlow, SDN Controller, Heuristic Algorithm, Switches, controller Imbalance, Delay, Jain’s Fairness Index