Minutes of the RTFM WG, Washington DC IETF, Wed 10 Dec 97

Reported by Cyndi Mills

Scott Bradner discussed multicasting and its associated management and
measurement issues, as set out in draft-bradner-multicast-problem-00.txt.
Important issues include scaling, making multicast tools mirror their
unicast predecessors, and making sure that multicast information is both
pertinent and easily understood. 

Sig Handelman reviewed the 'New Attributes' I-D, and asked for more
input from the mailing list as to what further attributes should be
considered, and how the new attributes should be implemented. 

Nevil Brownlee explained how he had implemented the 'Distribution-
Valued' attributes in NeTraMet, illustrating this with plots of
TurnaroundTime for SNMP packets.  It was suggested that two-dimensional
distributions would also be interesting. 

Nevil presented a proposal for using TCP sequence numbers as an
indicator of the quality of a TCP stream.  This was discussed;
consensus was that the total number of TCP bytes transferred would be
a useful RTFM attribute, but first and maximum sequence numbers would
not.  'Skip' and 'retransmission' counters would also be useful, but
these could not be aggregated. 

Ian Graham and John Cleary of The University of Waikato reported via
Nevil that they are measuring IP transit times (with real-time live
transit time histograms across the Internet) using ATM.  For details
see http://atm.cs.waikato.ac.nz. 

The IBM meter continues implementation of the RFCs and new I-Ds.  They
intend to make it available in some fashion in 1998. 

Version 4.1.0 of NeTraMet was released on 26 Nov 97 and is fully RFC
(and I-D) compliant.  The NeTraMet 4.1 source files include 32-bit
OC3MON source; OC3MON includes a NeTraMet meter as one of its analysis
modules.  New work includes a meter using input from Cisco NetFlow to
observe packets.  Some implications of this were discussed. 

NetFlow reports terminated flows, rather than being able to check on
flow status and "watching" current flows.  With NetFlow running in a
router you get source and destination AS numbers, which NeTraMet
doesn't otherwise collect.  NeTraMet cannot control what flows NetFlow
reports, but can perform data reduction of them. 

The WG reviewed its Goals & Milestones.  The consensus of the meeting
was that the Architecture and Meter MIB I-Ds need no further changes.
A revised version of the New Attributes I-D will be published shortly.
The WG's Goals and Milestones were revised as follows:

January 98  WG Last Call for Architecture and Meter MIB I-Ds.
            Submit to IESG for consideration as Proposed Standard RFCs.

April 98    WG Last Call for New Attributes I-D.  Submit to IESG
            for consideration as Informational RFCs.