Dynamic Load Balancing in Software Defined Networks
Translated title
Dynamisk Load Balancering i Software Defined Networks
Author
Boada Navarro, Marti
Term
4. term
Education
Publication year
2014
Abstract
This thesis investigates how Software-Defined Networking (SDN) can be used to dynamically distribute traffic across multiple possible paths to improve network resource utilization without overprovisioning. Building on SDN’s separation of control and data planes, it designs and implements a dynamic load-balancing algorithm, inspired by MPLS traffic engineering principles, that selects and reroutes paths based on link load. The algorithm is implemented in the Floodlight controller using OpenFlow and evaluated in an emulated Mininet topology. The work details data structures, new-flow routing, and rerouting mechanisms, and assesses behavior under both UDP and TCP traffic using metrics such as throughput, latency, jitter, packet loss, and link utilization. The excerpt does not provide quantitative findings, but the thesis presents the design, implementation, and emulation-based evaluation of the proposed approach.
Dette speciale adresserer, hvordan Software-Defined Networking (SDN) kan udnyttes til dynamisk at fordele trafik over flere mulige ruter for at forbedre udnyttelsen af netværksressourcer uden overdimensionering. Med udgangspunkt i adskillelsen af kontrol- og dataplan i SDN formuleres og implementeres en dynamisk load balancing-algoritme, inspireret af principper fra MPLS-trafikstyring, der vælger og omfordeler ruter baseret på linkbelastning. Algoritmen er realiseret i Floodlight-controlleren med OpenFlow og afprøvet i den emulerede Mininet-topologi. Arbejdet beskriver datastrukturer, rutning af nye flows og mekanismer til omruting, og evaluerer adfærd under både UDP- og TCP-trafik med metrikker som gennemstrømning, latenstid, jitter, pakkegtab og linkudnyttelse. Uddraget indeholder ikke specifikke kvantitative resultater, men specialet præsenterer design, implementering og emulationsbaseret evaluering af den foreslåede tilgang.
[This apstract has been generated with the help of AI directly from the project full text]