Extreme-value analysis of teletraffic data

An empirically veri1ed characteristic of the expanding area of Internet is the long tailness of phenomena such as cpu time to complete a job, call holding times, 1les lengths requested, inter-arrival times and so on. Extreme values of the above quantities are liable to cause problems to the e'cient operation of the network and call for e7ective design and management. Extreme-value analysis is an area of statistical analysis particularly concerned with the systematic study of extremes, providing useful insight to 1elds where extreme values are probable to occur and have detrimental e7ects, as is the case of teletra'cs. In this paper, we illustrate the main elements of this analysis and proceed to a detailed application of extreme-value analysis concepts to a speci1c teletra'c data-set. This analysis veri1es, too, the existence of long tails in the data. c © 2003 Elsevier B.V. All rights reserved.