4683991_monash_131687.pdf (4.68 MB)
Contributions to solving backbone VLAN identifier (BVID) problem and optimizing control-plane traffic in service provider transport networks
thesisposted on 2017-02-23, 01:42 authored by Bhamare, Deval
Internet traffic has grown at the rate of 40-100% per year and is only expected to grow at the same rate or more aggressively in the future. Due to the enormous growth in IP traffic volume, network operators are confronted by the unprecedented challenge of accommodating the IP traffic volume in already deployed networks in a short time-span. In addition to this, network operators are observing stringent latency and reliability requirements. The overall effect of these problems has been sub-optimal utilization of network resources resulting in higher network-wide capital and operational expenditures, along with lesser revenues. Hence, there is a need to optimize network cost while accommodating maximum traffic volume in the network and simplify the existing Carrier Ethernet Networks. In this thesis, we analyze high-speed telecommunication networks, with a focus on cost and performance optimization of high-speed Carrier Ethernet Networks. The first problem we focus on is that of optimization, in which, we try to minimize the number of network identifiers in the routers used in high-speed networks. Since available identifiers in the hardware are limited (due to the limited number of bits allocated to the identifiers), and requests for connections in the networks are increasing, it is desirable that such identifiers are re-used so as to maximize the number of connection requests satisfied in a network. In addition, we try to minimize the cost of the contemporary high-speed networks so that the internet service providers can benefit from maximum profits by reducing Capital Expenditures (CAPEX) and Operational Expenditures (OPEX). We focus, specifically, on reducing the total cost of the network interfaces at different network layers, satisfying all the traffic demands. In this pursuit, we study multilayer optimization from the perspective of deploying services so as to reduce the CAPEX in provider networks. To this end, we propose a 3-layer network hierarchical model based on IP, OTN and DWDM technologies. The goal of this work is to ascertain the type of traffic and how much of the traffic would require to be routed through a particular layer of the network to reduce the total cost of ownership. In the latter part of the research, we seek to answer an interesting question—how does Carrier Ethernet perform for Virtual Machines (VM) migration in data-center and cloud environments? The question arises since VMs form the central processing entity in data-centers and are crucial to facilitating cloud computing environments as well as the recent enhancements in Carrier Ethernet, which can be used as an efficient transport mechanism in DC/Cloud environments. To this end, we perform an extensive simulation study measuring the performance of VM migration over both flavors of Carrier Ethernet—namely PBB-TE and MPLS-TP. Our study concludes with an examination of the feasibility of Carrier Ethernet as a transport technology in data-centers and clouds. To extend the optimization work in high-speed carrier networks, we also focus on the implementation of the Centralized Control Plane for managed networks and try to optimize the Control traffic in the network. The overlapping of the control plane and the data plane in contemporary Carrier Ethernet (CE) networks leads to fragile and unmanageable networks. CE networks using packet technologies ought to be more manageable, scalable and robust. We propose an approach of a Centralized Control Plane to overcome the problem mentioned earlier. However, the control traffic generated due to the interaction between the centralized control plane and network elements increases exponentially as the number of nodes in the network increase and adds to the latency in the data traffic flow. The problem of control traffic overhead in managed networks is analyzed using an appropriate simulation model. A scheme to divide the network into smaller sub-networks is proposed so that total control traffic is always below a certain threshold. An ILP model and a heuristic approach are presented to strengthen the claim. The work is divided into separate chapters, as outlined in the Introduction. Thesis submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy of the Indian Institute of Technology Bombay, India and Monash University, Australia.