Re: int. cartography progress report

[Bernhard Stockman SUNET <[e-mail]>]

Hi all explorers in the network jungle.

Having read the Internet Cartography progress report some
things comes into my mind. (I realize that a lot of work
already has been put into this project but maybe my ideas
could have some additional value).

Problem specifications:

1. As pointed out using MIB variables (already in MIB-II or
here invented) the big drawback is the problem of having
these vars up to date or even initialized as they are
not necessary for network functionality. It would be
a VERY distributed database fully dependent on every
single site running anything equipped with SNMP, which
is rapidly migrating into every possible network equipment. 
(Soon I maybe will be able to switch the light on/off using SNMP :-).

2. The problem of not only having a distributed database but
at the same time using a proxy agent for part of the MIB vars 
not needed for network functionality.

3. The problem of not wanting to access every single box just
to be able to draw a map and maybe not even wanting SNMP access 
to certain network elements.

4. Using network element transient MIB vars will make the
maps dependent on temporary changes in the network and thus
maybe not reflect the stable situation.

Suggestions:

The idea of minimizing network access and at the same time
being able to access accurate topology data makes me think
of some kind of centralized topology server. Thus expanding the 
proposed SNMP proxy agent not only to collect certain MIB vars
but to collect all MIB vars needed for network topology for example
within one AS. This topology server could at regular intervalls 
query a selected set of network element belonging to its AS. 
Having some kind of timeout hysteresis for transient MIB vars 
will take away the effects of temporary situations.

This would make it easy to retrieve topology information for a certain AS. 
It would also simplify the managemnet of the MIB-ish vars as they could 
be configured into this server at init.

To query one or a few topology servers to get necessary network
data will probably not be that waste of resources as when having to
query every connected network element. Using SNMP authentification 
in one or another form would make it possible to set up an access 
scheme for servers that should be allowed to query network elements 
if restrictions on this are necessary.

Network elements not (yet) equipped with SNMP would be possible
to at least have some knowlede of using the MIB-ish database only.
(And maybe exand the tolpology quering possibilities to inlcude 
not SNMP/MIB objects in the true sense of an SNMP proxy agent).

Concerning tools for displaying topology data. Combined with the 
geographical server from Tom Libert at Merit, a topology server could 
replace the static topology info used by the netmap prog to produce 
geographically correct network maps but now with the auto-discover
capability.

Regards,
*******************************************************************************
        ___________	
       / ________  /	Bernhard Stockman   
      / //------/ /	SUNET/NORDUnet Operation Group.		  
     / //______  /      Royal Institute of Technology
    /  ------// /       Drottning Kristinas Vag 37A     Phone:  [phone]
   / _______// /	S-100 44 Stockholm, Sweden.     Fax:    [phone]
  / --------- /		Email: [e-mail] (Internet)   
  ------------		       BOSS@SEARN    (BITNET)

*******************************************************************************