Internet Backbone Network Traffic in Sudan

No Thumbnail Available
Mohammed Osman, Hiba
Journal Title
Journal ISSN
Volume Title
The growing demand of Internet service in Sudan makes it important to have accurate traffic statistics in order to help in the planning and provision of Internet links. This thesis presents statistics of the Internet traffic for Users connected through the public frame relay network .It also presents some observations that may help in improvement of the service to be of a better quality and affordable cost. Measurements and statistics in this research showed that most of the Internet users are concentrated in Khartoum center and Khartoum south areas, and are allocated percentages of the total bandwidth between 43% for users in Khartoum center area and 37% for users in Khartoum south area. Statistics of bandwidth distribution per user category showed that 42.9 % of the bandwidth is allocated to Service providers. Statistics of the daily traffic variations showed that the Overall traffic has its peak hour at 2 pm and drops to its minimum at 6 a.m. it also showed that the mean of traffic intensity is twice its standard deviation and the variance is very high if related to the mean. Per weekday analysis of the traffic showed that the average traffic utilization per weekday for the aggregate links maintains the same level with slight drop on Fridays. From the study of traffic variations per user category it is observed that traffic generated by local service providers has a great impact on the overall traffic pattern. The percentage of traffic uploaded and downloaded by each of the international links implies that the routing policy used to load share the upload traffic between the international links resulted in an equal load sharing. And that the shortest path of 18 % of downloads traffic is through Emirate link. A straight line with a slope of 0.5 can represent the relation between download and upload traffic for aggregate links. The packet arrival model of the aggregate links do not follow Poisson distribution, and is proved to be self similar .The queuing delay increased sharply for utilization values greater than 55%, and maintains the same level for utilization values less than 53%.