Fitting Distribution Modeling and Traffic Policing on Real Live Metro-E Network Data

Abstract

This paper presents a traffic characterization and statistical analysis of WAN Metro-E campus network internet traffic, addressing network congestion and delay issues. The increase in internet usage on campus can lead to various challenges, including traffic bursts that can impact the quality of service (QoS) experienced by users. The method involves traffic characterization and traffic policing on the Metro-E 50Mbps campus network using Python. The analysis will define the distribution model and policing the network of real-time internet traffic data where real live data at 50Mbps/6.5MB present burst traffic. The best-fitted model is the log-normal distribution model, which has the highest MLE score of -2726. Policing on inbound bytes before and after with a threshold of 50Mbps/6.5MB was done and a total of 3422404.81MB buckets were created and a total of 219273.88MB buckets after policing with a reduced percentage at 99.36%. Comparing actual traffic under three different policing scenarios compared to P1 and P2, policing on P3 has the highest traffic filtering at 3197986.21MB and the largest number of bytes filled in the bucket with a reduced percentage to 90.66%. This research is significant for Wan Metro-E Network's future QoS bandwidth management control mechanisms and optimization of network traffic performance.