/nsfnet/statistics/INDEX.statistics 29 October 1995 Merit Network Information Center Services NIC.MERIT.EDU FTP.MERIT.EDU FTP.MICHNET.NET NIS.NSF.NET (35.1.1.48) Merit's Network Information Center host computer, accessible via anonynmous FTP, contains a wide array of information about the Internet, NSFNET, and MichNet. NSFNET statistics are reported, as available, at monthly intervals. The statistics include only that portion of the ANSNet backbone that receives direct NSFNET sponsorship. The primary tools used to collect the statistics are the SNMP and other software modelled on NNStat. Monthly reports are in sub-directories according to the year in which the data was collected. Prior to 1993, files in these sub-directories adhere to the naming convention of "tN-YYMM.TYPE", where "tN" = network (t1 or t3) "YY" = year "MM" = month "TYPE" = report type Since the T1 network was retired in late 1992, reports dated 1993 or later have the prefix "nsf" instead of "t3". The various report types are defined as: delay 1st, 2nd (median) and 3rd quartiles of one-way delay times between each NSFNET node on the backbone during the one- month reporting period. Prior to July, 1991, the minimum, maximum, and mean were reported. pnss Daily inpacket totals for each NSFNET node on the backbone during the one-month reporting period. These files are tab delimited for easy loading into spreadsheets. bnss Daily inbyte totals (in millions) for each NSFNET node on the backbone during the one-month reporting period. These files are tab delimited for easy loading into spreadsheets. pnets Traffic broken down by network, measured in packets, which traversed the backbone during the one-month reporting period. Only traffic that involves an NSFNET-sponsored network is included in this report. Entries for non-NSFNET-sponsored nets include only traffic exchanged with NSFNET-sponsored nets. bnets Traffic broken down by network, measured in bytes, which traversed the backbone during the one-month reporting period. Only traffic that involves an NSFNET-sponsored network is included in this report. Entries for non-NSFNET-sponsored nets include only traffic exchanged with NSFNET-sponsored nets. country Traffic broken down by country, measured in both packets and bytes, which traversed the NSFNET backbone service during the one-month reporting period. ports Traffic broken down by service, measured in both packets and bytes, which traversed the backbone during the one-month reporting period. highlights A one-page summary of the "ports" report, listing the top 30 and other selected services. For example, the file "nsf-9301.pnss" is a report of the January '93 packet counts of all NSFNET NSSs on the ANSNet Backbone. Other files available in this directory are: history.bytes Growth in traffic on the NSFNET backbone service as measured in bytes. Byte traffic measurements are available beginning March 1991. history.hosts Growth as reflected in the number of computers and domain names on the Internet as reported in Mark Lottor's Internet Domain Survey. history.netcount Growth as reflected in the number of domestic and foreign networks having announcement to the NSFNET infrastructures. history.packets Growth in traffic on the NSFNET backbone service as measured in packets. history.ports Distribution, by percentage, of most common uses. nets.by.country Distribution, by country, of networks announced to the NSFNET backbone service. nets.by.state Distribution, by state, of U.S. networks announced to the NSFNET backbone service. nsfnet-t1-stats.ps Traffic Characteristics of the T1 NSFNET Backbone. Kimberly Claffy, George Polyzos, and Hans-Werner Braun. PostScript, 11 pages, 428,978 bytes Jan 1993 nsfnet.retired Merit retires NSFNET Backbone Service. 65 lines, 3,070 bytes May 1995 restrict.nets Description of NSFNET route filtering policy. top10.transition.ps Chart showing traffic volume over NSFNET Backbone Service during the transition to the new NAP technology. PostScript, 1 page, 107,105 bytes Mar 1995 transition.status Status report of the transition from the NSFNET Backbone Service to the NAP architecture. 127 lines, 4,717 bytes Mar 1995 Notes about the statistics collection: The primary tools for data collection are SNMP and NNStat/ARTS. SNMP-based reports are history.packets, history.bytes, pnss, and bnss. NNStat/ARTS-based reports are history.ports, ports, highlights, pnets, bnets, and country. The SNMP and NNStat/ARTS measuring tools reside at somewhat different locations at the backbone nodes and use differing techniques to collect data. As such, they do not measure quite the same packet flow. Also, NNStat/ARTS data collection switched to a sampling method in September, 1991, as management of the sheer volume of data had become untenable. Late in 1991, NSFNET began the migration from a T1 to a T3 Backbone Service, with T1 traffic counts decreasing accordingly. By June, 1992, most of the traffic had transitioned to the T3 Backbone Service. The T1 pnets, bnets, country, and ports reports were discontinued at that time, since the remaining T1 traffic could no longer be considered representative of the NSFNET Backbone Service as a whole. Full statistics reporting was resumed in November, 1992, when ARTS, the completely redesigned and implemented NNStat replacement, was put into production. In the early days of NSFNET, every node on the backbone was an NSFNET node. When the T3 technology was implemented in 1992, NSFNET was moved to a larger network that had some nodes that did not belong to NSFNET. At this point, the "populations" being measured by the varying collection methods changed dramatically. The SNMP-based reports now report only the traffic that enters the NSFNET Backbone Service at an NSFNET sponsored node. The ARTS history.ports, ports, and highlights reports apply to the entire T3 backbone, including the non-NSFNET sponsored nodes. The ARTS pnets, bnets, and country reports, however, record the subset of traffic that either comes from or is destined to NSFNET sponsored networks, regardless of where the packets enter the T3 backbone. Partly for consistency reasons, Merit/NSFNET continues to interpret the SNMP-based packet reports as the truest measure of growth of the NSFNET Backbone Service. The traffic counts reported in these subdirectories again started to decrease during December 1994, when networks began to migrate to the new NSF network architecture. In the new architecture, traffic is exchanged at interconnection points called NAPs (Network Access Points.) On April 30, 1995, the NSFNET Backbone Service was successfully transitioned to the new network architecture. Although the reports are inclusive through the end of the NSFNET service, the November 1994 reports are the last to reflect the nature of the NSFNET backbone traffic in its entirety.