NSFnet Statistics Collection System NNStat Release 3.2 December 1992 (Maintenance release) Developed at USC Information Sciences Institute by Bob Braden and Annette DeSchon With contributions from: Dave Katz, Merit Phil Wood, LANL Alan Stebbens, UCSB Kannan Varadhan, OSU Michael Walker, NASA Jeff Mogul, DEC Carl Schaefer, Cisco Statspy is designed to run on: Sun-3s and Sun-4s, under Sun OS3.x and Sun OS4.x; Vax and MIPS architecture machines under Ultrix; IBM PC RTs. Tar file includes: User Document: doc.ms -- nroff -ms source doc.txt -- text CHANGES -- list of changes Source files: Makefile -- used to build statspy, rspy, collect, and lookupnames *.h, *.c -- source pfilt.h, pfiltrt.h -- one of these should be useful to RT installation Example/suggested configuration and parameter files: parm.sample* -- suggested configuration files for statspy parm.enum, parm.portenum -- 'include' enum files. networks.txt -- table used by lookupnames (from standard hosts.txt) Scripts and awk files for data reduction: count-totals.alias -- script to define count-totals command bin-totals.alias -- script to define bin-totals command dailytotals.awk, dailytotalsB.awk, newlines.awk, subtotals.awk -- AWK scripts used by count-totals, bin-totals To make programs: comment in the appropriate set of Make macros (see Makefile) and do: make all Note that can generate more efficient version without subnet virtual fields, with: make all SUBNET= To run statspy on a SunOs or IBM RT system, you must first become super-user. To run statspy on an Ultrix system, you need not be super-user, but the super-user must first enable the use of promiscuous mode using /etc/pfconfig (e.g., "/etc/pfconfig +p -a"). To start statspy, execute the shell command: statspy parm.sample1 Sun OS/4.0.x contained a serious bug in NIT. If this is not fixed, statspy will fail immediately with an error message: "NNStat: Fatal NIT error Statspy depends upon bug fix in nit_if.c" Fixed modules for Sun3 and Sun4 are available for anonymous FTP from host venera.isi.edu with pathnames: pub/nit_if.o pub/nit_if.o.sparc To see what statspy is collecting, enter the command line: read * To see the commands, enter ? If you don't want to tie up a window, however, start it as a background job: statspy parm.sample1 >& statspy.log & To start collect, execute the shell command (Superuser not necessary!): collect -e enum.NSF -i 5 -c 60 -h '*' Here is domain name or dotted-decimal number of the host on which statspy is running. This will poll for data every 5 minutes, saving checkpoints every 60 minutes, and put the data into a set of files, one per object. To run collect in background, putting standard error into a file, use: collect -e parm.enum.collect.sample -i 5 -c 60 -h '*' >& error.log & To install the reduction commands, issue: source count-totals.alias source bin-totals.alias These will issue alias commands to define count-totals and bin-totals for the current session. It may be desirable to edit your .login file to include these alias definitions. To summarize the Total Count data in all log files whose names include the string "IP", use: count-totals v=1 *IP* for hourly totals, or for daily totals: count-totals *IP* To summarize by corresponding bins, use: bin-totals *IP* To remotely control statspy, use: rspy -h parm.rspy where specifies the network host on which statspy is running. _________________________ For more info, READ THE DOCUMENT! We also suggest that you might want to get on the mailing list bytecounters@ISI.EDU. This list was set up to report problems and seek solutions on statspy, or to discuss other issues related to gathering usage statistics. To get onto the list, send a request message to bytecounters-request@ISI.EDU. Questions and comments to: braden@isi.edu deschon@isi.edu