CURRENT_MEETING_REPORT_
Reported by Kenneth Goodwin/PSC
NJM Minutes
Instead of putting actual copies of maps into central repositories,
pointers as to where to get the maps should be placed there. This would
make it easier and quicker to update maps, since the network manager
does not have to distribute maps, just update the local copy.
It seems as though the very bade traceroute has disappeared from many
peoples' list of tools. Matt Mathis said that it appears that
traceroute is included in Ultrix 4.2, although it is unknown whether it
does third party or not. Sources for traceroute are available via
anonymous ftp from ftp.lbl.gov (may not include the third party mods)
and from nic.near.net (in the pub or src directory).
Also, Matt M. provided some examples of traceroutes into the backbone
that failed. The common denominator being that if the endpoint or the
corner of the third party traceroute is in the backbone, it will fail.
Merit says that this may be by design.
Merit/ANS discussed their architecture plans for the CNSS [see also
slides from the plenary technical presentation by Elise Gerich and
Jordan Becker - EFH]. Currently, they decrement the TTL through the
routers, but would like to make the CNSS's in a POP appear to be one
router by only decrementing the TTL on entry or exit. This would make
the separate CNSS's invisible. One point against this is that in a
tightly coupled FDDI, this could create a giant loop. (??) Merit also
said that future cards will receive routing updates from the CPU.
Fallback paths were also discussed. (In particular fallback paths of
lower capacity) Some important points were that if a backup path exists,
then it should be carrying some traffic all the time, so that the backup
path's status is known at all times. Backup paths of lesser capacity
can easily be flooded by production traffic, so some means of limiting
traffic must be made. Things like mail can be MX'ed to the backup path,
and all other traffic could be blackholed.
As nets become faster, monitoring is becoming harder. Merit is
currently using periodic sampling of 1 in every 50 packets on the T3.
Merit would like to use a more stochastic process. More work needs to
be put into this to determine a more accurate sampling process.
With more and more nets appearing, some better method of reporting
outages is needed. An outage of a campus with large nets could easily
flood people. A template for reporting outages is needed, so that a
database can parse these messages and store the information. Thus, one
need not even read the mail, but query the database for the net in
question.
Merit is working on something like this for trouble ticket tracking.
1
Since high schools are entering the internet, two problems are
occurring. Under what name should the schools appear. (us or edu)
Also, how do we get them started? Nearnet offers a full service option
for new people, that completely orients the newcomer. User Services
should also target the end user and not just the network operators.
Attendees
Vikas Aggarwal vikas@JVNC.net
Jordan Becker becker@nis.ans.net
Eric Carroll eric@utcs.utoronto.ca
Henry Clark henryc@oar.net
Tom Easterday tom@cic.net
Susan Estrada Estradas@cerf.net
Vince Fuller vaf@stanford.edu
Maria Gallagher maria@nsipo.arc.nasa.gov
Kenneth Goodwin goodwin@psc.edu
Jack Hahn hahn@umd5.umd.edu
Eugene Hastings hastings@psc.edu
Ittai Hershman ittai@nis.ans.net
Daniel Long long@nic.near.net
Matt Mathis mathis@psc.edu
Philippe Park ppark@bbn.com
Marsha Perrott mlp+@andrew.cmu.edu
Robert Reschly reschly@brl.mil
Ron Roberts roberts@jessica.stanford.edu
Timothy Salo tjs@msc.edu
Tom Sandoski tom@concert.net
Roxanne Streeter streeter@nsipo.nasa.gov
Ross Veach rrv@uiuc.edu
Chris Waters-Pierandozzi waters@jvnc.net
Gerard White ger@concord.com
Cathy Wittbrodt cjw@nersc.gov
2