Your browser does not support JavaScript!

Home    Measurement-driven modeling of traffic demand and short-term forecasting in large WLANs  

Results - Details

Add to Basket
[Add to Basket]
Identifier uch.csd.msc//2007raftopoulos
Title Measurement-driven modeling of traffic demand and short-term forecasting in large WLANs
Alternative Title Μοντελοποίηση δικτυακής κίνησης βάσει πραγματικών μετρήσεων και βραχυπρόθεσμη πρόβλεψη φόρτου σε ασύρματα τοπικά δίκτυα
Author Raftopoulos, Elias
Thesis advisor Παπαδοπούλη, Μαρία
Abstract Wireless Local Area Networks have seen enormous success in response to the growing demand for wireless access. In this context network operations such as capacity planning, admission control, and load balancing, will become more relevant and will have to be properly engineered to match the WLAN constraints. It is of critical importance to understand the performance and workload of the wireless networks and develop wireless networks that are more robust, easier to manage and scale, and able to utilize scarce resources more efficiently. Thus, it is important to perform empirical studies to measure the phenomena of interest in real-life networks and formulate realistic models of user communication and association patterns. This can be beneficial in the administration and deployment of wireless infrastructures, protocol design for wireless applications and services, and their performance analysis. Moreover, the development of testbeds, tools, benchmarks, and models is of tremendous importance and can motivate further performance analysis and simulations. In this work, we study a large infrastructure and model the traffic load at APs. We take advantage of the wireless infrastructure of the University of North Carolina (U.N.C.) to draw large amounts of different types of measurement data. We then exploit the spatial and temporal resolution available in data traces in deriving models for traffic demand. Our contributions have a strong methodological element. For example, in modeling traffic demand we adopt a hierarchical framework that is found to capture demand in various levels of spatial scales, ranging from individual buildings to groups of buildings and network-wide. Throughout this work, we make heavy use of statistical tools; clustering techniques help us address scalability concerns in traffic demand modeling. Based on the modeling methodology, we design forecasting algorithms to predict the traffic load at APs in different time-scales. We then apply these forecasting algorithms on real traffic traces acquired from the most heavily utilized APs and evaluate their performance.
Language Greek
Issue date 2007-09-21
Date available 2007-10-24
Collection   School/Department--School of Sciences and Engineering--Department of Computer Science--Post-graduate theses
  Type of Work--Post-graduate theses
Views 385

Digital Documents
No preview available

Download document
View document
Views : 0