Please use this identifier to cite or link to this item: http://prr.hec.gov.pk/jspui/handle/123456789/10406
Title: On the Design and Optimization of Next Generation Virutal LANs
Authors: Hameed, Abdul
Keywords: Computer Science
Issue Date: 2018
Publisher: National University of Computing and Emerging Sciences, Islamabad
Abstract: Virtual Local Area Networks (VLANs) provides logical grouping in LANs that nd many applications. A little research work has been reported for designing optimal VLANs topologies. An optimal VLANs topology for a given network is one of the best topology, among all possible topologies, considering the purpose of VLANs implementation. Here we propose the idea of Next Generation (NGN) VLANs that have the best possible topology implemented in the network. An additional feature of the NGN VLANs is that it allows runtime con guration changes to its topology for coping with the stringent requirements of dynamic networks like cloud datacenter, LAN party networks etc. Such a dynamic runtime con guration is not possible with the existing manually con gured VLANs. Since VLANs topology optimization is an NP-complete problem, we therefore proposed both heuristics and metaheuristics approaches for the optimization of NGN VLANs. There can be many goals of the optimization, the example objectives considered in this research work are maximizing tra c localization and enforcing the prede ned group membership policy. First, we proposed a heuristic named simple set-based (SS) algorithm which uses greedy searching for optimizing VLANs topology with the objective of tra c localization. The SS algorithm utilizes the tra c statistics of the network and proposes a grouping layout for the network nodes. In such a layout, nodes exchanging large amount of tra c are consolidated into the same group. A graph-based optimization technique is next proposed for nding strong components which represents a group of nodes extensively communicating with each others. Genetic algorithm (GA) is a well-known nature-inspired metaheuristic tool for solving combinatorial optimization problems. A GA-based solution is proposed for maximizing tra c localization of the VLANs design. The multiobjective version of the problem with the objective of both maximizing tra c localization and enforcing prede ned group membership policy is tackled with state of the art multiobjective optimization algorithms. These algorithms are PAES, SMPSO, OMOPSO, NSGAII, e-NSGAII, NSGAIII, MOEA/D, e-MOEA and GDE3. The order of best algorithms found is SMPSO, MOEA/D and GDE3 respectively. The second feature of NGN VLANs i.e. allowing runtime con guration of its topology is employed with software-de ned networking (SDN). First, a detailed review is carried out to identify the functionality overlap between SDN and VLANs as both technologies has some features in common. With Floodlight SDN controller, we are able to make runtime con guration changes to VLANs, both in an emulated and in real SDN testbed. The proposed framework for runtime con guration of VLANs is customized for maximizing the tra c localization where it reacts to continuous spikes in the inter-VLANs tra c. It nds an updated VLANs topology that can better localize the current tra c trends. The recon- guration of the VLANs topology is carried out in matter of seconds without producing any disruption in the ongoing communication sessions in the network. With NGN VLANs, the network administrator will be able to decide an optimized VLANs structure thus maximizing the bene ts of VLANs implementation. With runtime con guration feature, clouds applications will be able to utilize existing legacy VLANs capable switches thus saving capital investment in the existing infrastructure.
Gov't Doc #: 17647
URI: http://prr.hec.gov.pk/jspui/handle/123456789/10406
Appears in Collections:PhD Thesis of All Public / Private Sector Universities / DAIs.

Files in This Item:
File Description SizeFormat 
Abdul Hameed_CS_2018_FAST NU_PRR.pdf3.99 MBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.