Path: utzoo!attcan!uunet!munnari.oz.au!jatz.aarnet.edu.au!pte900
From: pte900@jatz.aarnet.edu.auP (Peter Elford)
Newsgroups: comp.dcom.sys.cisco
Subject: Re: IP Accounting via SNMP
Message-ID: <1990Sep28.094936@jatz.aarnet.edu.au>
Date: 27 Sep 90 13:19:36 GMT
References: <26881@boulder.Colorado.EDU> <90.269.16:58:45@ira.uka.de>
Reply-To: pte900@jatz.aarnet.edu.au (Peter Elford)
Organization: AARNet
Lines: 46
Nntp-Posting-Host: jatz.aarnet.edu.au


In article <90.269.16:58:45@ira.uka.de>, nipper@iramu1.ira.uka.de
(Arnold Nipper) writes:
|> How long does it take to let's say retrieve ~1000 entries
|> from the CISCO via SNMP. We retrieve the data via a telnet
|> program written by Daniel Karrenberg (dfk@cwi.nl) and this
|> works quite fine. With another tool you can have domainnames
|> or information about the IP-networks given by NIC/NSF/RIPE
|> instead of pure IP-addresses. This tool was written by Daniel too.

This is true. Because of the MIB structure, each IP accounting table 
entry (four SNMP variable instances) requires a separate SNMP query.
They packets won't be very large, so the time to retrieve "~1000" entries
is a function of your network speed and the ability of the router to
respondto the queries. Pulling it pack through a telnet sessioon will
be more efficient because MIB I defines inherently scalar objects only;
the telnet session gets blocks of data (equal to the screen size) thus
giving you (sort of) portions of vectors of information.

This limitation (of SNMP) is being addressed by the IETF: take a look
at draft-ietf-snmp-mibdefinitions-00.txt, which talks about columnar
objects.

|> >it's worth having :-),

You have misquoted me. I was asking sites elsewhere in the Internet to make
it available for ftp only if they thought it was worth having.

|> ... because it's via SNMP???

Well yes! If you have only a small IP accounting table this is a more
elegant way of doing it. Sure, for 1000+ entries elegance is less important
compared to getting the data, but if people don't make use of new facilities
like SNMP then they are not going to develop to be useful (like the 
columnar objects).

I did forget to mention the responses I got from the guys who had done
telnet based IP accounting grabbers (there were two), for which I 
apologise.

Peter Elford,                           	e-mail: P.Elford@aarnet.edu.au
Network Co-ordinator,	 			phone: +61 6 249 3542
Australian Academic Research Network,		fax: +61 6 247 3425
c/o, Computer Services Centre,			post: PO Box 4
Australian National University			      Canberra 2601
Canberra, AUSTRALIA