Electronic Thesis and Dissertation Repository

Thesis Format

Monograph

Degree

Master of Science

Program

Computer Science

Supervisor

Haque, Anwar

Abstract

Accurate network costing provides insightful information to any ISP for better network planning, profits, and decision making. Developing precise cost models for communication network services has always been a challenge for Internet Service Providers (ISP) due to the complex nature of today’s advanced shared cloud and network infrastructure. Currently, developing and maintaining such cost models require significant effort and time for the network planners in an ISP. The proposed novel methodology reduces the development cycle time significantly for the cost model, which leads to the ISP’s operational cost savings. We also experimented with K-means clustering for grouping router costs in the study, which provided similar unit cost results. To prove the operational savings, we evaluated a quantitative example considering the current practice as well as our proposed methods. We considered three network services: IPVPN service, Transport Lease service, and High-Speed Internet service for the experiments. We conducted simulations, and estimated service unit costs to validate the accuracy and effectiveness of our proposed approaches. We have compared results from proposed strategies with the existing cost mechanism and computed the performance improvement cost gap for different network sizes. This cost gap (delta) exhibited that the difference between the service cost values is significantly negligible, which proved the efficiency of our cost model.

Summary for Lay Audience

There is a significant increase in communication services from the past several years. Service providers need to manage their network as well as their associated costs. For this purpose, a sustainable cost model facilitates them to evaluate decisions for their network. This thesis aims to present a novel approach in developing a cost model for network services that helps the network planners and modelers that would result in potential operational savings.

The thesis briefly explained the notion of network costing and investigated several methodologies related to network costs, which we classified into wireline and wireless networks. Our primary focus is to explore and analyze service cost modeling methodology from an ISP perspective. We first investigated the mechanism that already being used by the researchers and industry and then proposed two service cost modeling methodologies for grouping router cost values. We analyze our approach using publicly available router data. We implemented the service cost model framework, in which the computation involved cost and capacities data of the relevant network nodes/routers. The first method is the simple mathematical approach of grouping the router’s cost and capacity values into three types comprised of minimum, median, and maximum values. In the second method, we applied the K-means clustering algorithm that grouped the data into three router types based on the cost values similarity. This thesis considered three network services: IPVPN service, Transport Lease service TLS, and High-speed internet service HSI for various network sizes for unit cost estimation applying both approaches.

The service unit cost results obtained from the service cost model framework for all three services considered. We also compared the unit cost results estimated from the current cost method practiced in the industry with the unit costs values achieved from the proposed methods. The purpose of the comparison is to demonstrate the closeness of service cost results. Also, to validate the accuracy of the results obtained from the proposed cost modeling methodologies that would assist in saving ISP Opex by reducing the employee hours spent to model service costs.

Share

COinS