Internet Info 1994 March/Internet Info CD-ROM (Walnut Creek) (March 1994).iso/inet/ietf/opstat/opstat-minutes-92jul.txt

Text File | 1993-02-17 | 8KB | 200 lines

Editor's Note:  Minutes received 7/21

CURRENT_MEETING_REPORT_

Reported by Bernhard Stockman/SUNET

Minutes of the Operational Statistics Working Group (OPSTAT)

Agenda


   o Administrative Items
   o Review Internet Draft, is it ready to progress to RFC?
   o Development of tools
   o Review Charter, Goals, and Milestones


Review of Comments Received About Internet Draft

Several comments were made by various Working Group participants.  Jon
Boone (PSC) wants stop-time and filename in a header section so that it
isn't necessary to scan the entire file to find the ending time of the
data.  Agreement was reached to change the time_section to label_
section as follows:



     label_section ::= ``BEGIN_LABEL'' <FS>
     <start_time> <FS>
     <stop_time> <FS>
     <data_file> <FS>
     ``END_LABEL'' <FS>



There was a question about file setup...  is there a need for one big
file or lots of little files?  Should there be multiple sections within
a file?  Matt (PSC) noted that important information was not mentioned
and could be potentially confusing.  They wanted to use very large files
because of their tape storage facilities.  Agreement was reached to add
a sentence saying the specifics of how files are physically arranged is
outside the scope of the document.

There was some confusion about the use of tags and variables in the
poll-data section.  Are there multiple tags for different sets of
variables?  It was noted that the draft is vague on time aggregation in
this context and there was no clear way to do it.  There is a need to
provide a representation for, say, the average for 1 hour, 1 minute, and
a maximum value, as well as, a need for classes of operators since
aggregation is different for counters vs.  an interface status variable.
Agreement was reached to rewrite the section to make it more clear.

Would the addition of comments within the data files be useful?  Yes,
add comments, something like:

                                   1





     FS :== ``,'' | <LF> | <LF> # text <LF>


Additional questions were posed by Dave S. (BNL).


   o What is the use of the networkname field?  Ross noted that network
     names were unique.  Evan noted that the sharing of data among
     networks would be facilitated such that data wouldn't possibly
     become confused.  Consensus was reached that the networkname field
     was useful :-).

   o Question about the routername.  This brought up a bigger discussion
     about how addresses are bound to each interface on a router.
     Usually the name of a router is tied to the interface most commonly
     used to access it.  The name must be unique.  Discussion digressed
     to involve yellow post-its and neon lights to name routers.  Matt
     mentioned limitations of the DNS to name routers usefully.  No
     action was taken.

   o Questions about the linkname prompted discussion on what should be
     in it and what would be mandatory.  The concept of a virtual link
     is needed to represent one or more physical links which can be
     grouped for statistical purposes.  Should the name represent the
     ISO layer 2 or layer 3 name?  Should there be an external name map
     to map the linkname field to a meaningful string?  How should
     information be encoded in the field?  Should there be resource vs.
     time aggregation?  There are several unanswered question which
     could be answered in a later document covering implementation
     details.


Ed Reeder (IBM) suggested several editorial comments.  All were approved
as suggested.

In Section 5.1 there was a question about the difference between the raw
data and the presented data.  Agreement was reached to rewrite
sentence/section to make more clear about the difference between the
two.

In a separate discussion, Matt suggested adding a field to show
specifically whether the data had been aggregated, versus having an
implicit indication currently.  Everyone agreed to this change.

There was some question about having a minimum value as well.  Lengthy
discussion about the minimum value always being very close to or equal
to 0.  Consensus had been reached at an earlier meeting to drop the
minimum value.

James Barr (NIKHEH-H) asked a question about adding a comment.  Since
this had been agreed to previously, no further discussion was held.

Peter Fenwick (Univ.  of Auckland) pointed out a syntax error in Section

                                   2





6.1.3 in the data field specification where the number of ``[`` and ''
]'' were unequal.  Everyone agreed the error should be fixed.

Pietrak Rafal had extensive comments on the aggregation periods in the
draft.  Comments about the effect of extra weekend days in a month
skewing data for a physical month, as well as, as questions about the
period of time we wanted a peak value for raised.  Everyone agreed that
two hours was too long of a period.  Agreement was reached to leave the
time values as they were in the Internet Draft.

Is the draft ready for RFC'ing?

Agreement was reached that the draft was getting close and needed the
changes mentioned above.  Once the changes are made, Bernhard will send
a copy for review before forwarding it to the IESG.

Future Tool Development

Eric Hood, Executive Director of FARNET, made some comments about
network statistics and the availability of financial assistance from
FARNET to help fund some amount of development.  Eric was going to have
the FARNET staff survey networks to see what tools are currently
available and what is under development.  Everyone agreed that there
needed to be reference implementations to flesh bugs out of the draft
(and future drafts) and show the direction of future work.  A Consensus
was reached on the need for a common way to store data and to share
common tools which are freely redistributable (in the public domain).
We are currently unsure of the questions which will need to be asked at
some point about what statistics are important and implementations will
help answer the questions.

Eric will do the survey and forward the results to the OPSTAT Working
Group.  Discussions about tools will continue at the Washington, DC
IETF, November 16-20, 1992.

Attendees

Vikas Aggarwal           [e-mail]
Tony Bates               [e-mail]
Henry Clark              [e-mail]
Robert Collet            [e-mail]
Scott Cossette           [e-mail]
Osmund de Souza          [e-mail]
Tom Easterday            [e-mail]
Roger Fajman             [e-mail]
Stefan Fassbender        [e-mail]
Cliff Frost              [e-mail]
Martyne Hallgren         [e-mail]
Eugene Hastings          [e-mail]
Eric Hood                [e-mail]
John Labbe               [e-mail]
Walter Lazear            [e-mail]
Hock-Koon Lim            [e-mail]
Kim Long                 [e-mail]
Matt Mathis              [e-mail]

                                   3
^L




John McKenna             [e-mail]
Chris Myers              [e-mail]
Kraig Owen               [e-mail]
Sonya Reimer             [e-mail]
Robert Reschly           [e-mail]
Bernhard Stockman        [e-mail]
Ross Veach               [e-mail]
Curtis Villamizar        [e-mail]
Evan Wetstone            [e-mail]
Chris Wheeler            [e-mail]
Linda Winkler            [e-mail]



                                   4