Draft NETSTAT Minutes for 7 Dec. 1995
hastings at psc.edu
Sat Jan 14 22:33:40 UTC 1995
Network Status Reports met on 7 December, 1994 from 0930 to 1145.
Scheduled presentations were ANS (Rob Lehman), NSFNET Transition
(Elise Gerich), InternetMCI (Phill Gross), BBN (John Curran), Sprint
NAP (Tim Clifford) and SprintLink (Sean Doran)
In addition to the presentations and Q&A, several meta-issues were
raised or implied (furished along with my attempt to digest the pulse
of the group).
Is it appropriate to give what is essentially a marketing
presentation at IETF? The consensus seems to be "no" if it is
solely and blatantly such, but there is dissention as to where
the line is.
With the increasing activity of for profit concerns in the
Internet, will it be possible to continue the openness of past
years' presentations wrt traffic, performance, uptime
statistics, etc.? [Despite the fears or predictions of some,
there appeared to be some continuing openness, but also examples of
extremely closed practices. I think that collegial backpressure
has ceased to be sufficient to maintain the historical levels
of disclosure and cooperation. - efh]
In the light of these questions, what is the charter of this
group? (including what is desirable vs. what is acceptable). [I
think this is an ongoing process. Does commercial necessarily
mean concealed? I think in the context of an interconnected,
interdependent environment, it can't in the long run. - efh]
Beyond concerns about openness, there are areas of concern not
being addressed at all: What is the forum for operations,
engineering and troubleshooting above TCP/UDP? Traditional IP
regionals and carriers have been content to (determined to?)
focus their energies on transport and routing. What is the
forum for general end-to-end problem solving?
ANS Rob Lehman,<rll at ans.net>
[See slides - only supplemental notes follow]
ANSNET traffic surpassed 100m inbound packets in Nov.!
CIDR note: after AUP disappears, there ought to be further
ANS' NAP connectivity status:
Connection to the Sprint NAP was installed on 9-21-94. At this
time (7 Dec. 94) it's only been operational since 2 weeks, but
substantial traffic has been exchanged
Connection to the PacBell NAP was installed on 10-14-94, but is
not in production.
Connection to the Ameritech NAP was installed on 11-22-94, but is
not in production.
A MAE+ FDDI Installation plan has been established. Its exact
schedule is contingent on logistics. Real Soon Now.
ANS did link optimizations for cross-country trunking, to address an
imbalance in the relative utilizations of their northern and
southern routes. After testing, they are further considering
deployment of Random Early Drop.
Transition Elise Gerich <epg at merit.edu>
[No slides - see contemporaneous presentations to IESG, etc.]
NONE of the ENSSen have been retired. The regionals' transition to
non-NSF Inter-Regional Connectivity has been slow. Target dates of
Nov. 1 and Dec. 1 have been missed, Merit is hoping for ~ Jan. 1
terminations + 60 days...
THEnet and MOREnet have made the transition to SprintLink. SURAnet
has moved traffic to MCInet, but has not formally notified Merit of
the ability to terminate the ENSS connection.
CA*NET is close to transition to MCI
THEnet is sharing Inter-Regional Connectivity with Sesquinet
Interconnection point status:
The Sprint NAP has connections from Sprint, MCI, and NSFNET.
In DC, at MFS facilities, everyone is on MAE and is committed to
connecting to MAE+.
At the Ameritech NAP, MCI and NSFNET are about to start peering,
with Sprint to connect soon.
PacBell has MCI and ANS are peering, and Sprint will connect
Route servers are present at MAE, the PacBell NAP, the Ameritech
and the Sprint NAP.
Routing DB deployment and transition. The applications are moving
from PRDB (@merit) to RA (db-admin at radb.ra.net). At some point,
Merit will do dual use of NACR and RADB forms, followed by
retirement of NACR (transition ca. Jan. 15.)
Merit has recommended to NSF that during transition (Jan.-Jun.,
there should be no need to specify AUP.
If anyone is still relying on PRDB reports they should contact
InternetMCI Phill Gross <0006423401 at mcimail.com>
[see pretty slides. In color. Phill apologized.]
Problems were experienced transitioning SURAnet, thus delaying the
CoREN schedule. "DS3 networking is not yet a commodity service." -
experienced some problems in the routers.
Questions from the floor:
What management platform does InternetMCI use? Phill: Hewlett-
Packard OpenView, plus homegrown tools and extensions.
What Other nets beside CoREN is MCI serving? Phill: CA*NET is
partially transitioned...(at 3Mb), WIDE (at T1) and BTnet (at
InternetMCI NAP connectivity status: Chicago (Ameritech) was
connected a the end of last week, MCI is connected and peering at
Sprint, MAE and CIX. ANS and MCI are interconnected at FDDI at
Hayward CA. common PoP (Phill: Thanks for the cooperation, ANS!)
Q: When will MCI be connected to FIX-W? Phill: I don't know. It's
underway, maybe by the end of year, approximately the same as CIX-
Q: What's the status of the vBNS? Phill: we're deploying a testnet,
and full rollout 1Q95 at OC-3. The testnet should be operational
BBN John Curran <curran at bbn.com>
[Admittedly marketing slides.]
BBN is upgrading the NEARNET spine in the Boston area, migrating
from microwave Ethernet to fiber (MFS 10Mb over NYNEX T3).
Currently BBN has almost 2000 SNMP managed items...
Operations/NOC/NIC tip: you can head off phone calls by giving
BBN is now offering "Turnkey Internet Server" (Pentium/BSDi)
Internet Site Patrol - managed firewall (BBN CONTROLS it), with
remote management done by BBN ISC over a secure channel. It's
derived from TIS products, and supports telnet, ftp, smtp, nntp,
BBN strenuously wants to know how to pursue end-to-end problem
solving (they still subscribe to 'the router is BBN's, we control
the horizontal, we control the vertical).
Sprint NAP, Tim Clifford <tcliff at sprint.net>(New PI)
Report on the Sprint "NY" NAP (in Southern NJ). It is now (Dec. 94)
a dual FDDI ring, and in Jan. will be converted to a DEC Gigaswitch.
[See before and after diagrams]. Connected, parties include MCI,
NSFNET, Sprint and Cerfnet. A Route Server is present and running.
Aside: Sprint has "always thought about ATM" [the implication being
that they felt is wasn't ready - efh]
Question: the DEC GigaSwitch has DS3 cards. Have you considered
access this way? Tim: not really. [It's my understanding that a GS
ATM interface can only talk to other GS. - efh]
Q: being already co-located, can folks try alternate interconnect
technologies? Tim: likely. Try us.
SprintLink, Sean Doran <smd at sprint.net>
Sprintlink's T1 backbone is melting down. They're converting to DS-
3s or parallel T1s in the next few weeks.[see diagram]. T3 customers
feed directly into BB routers to avoid saturating local FDDI rings
For trans-US international connectivity, ICM has 2 T3 between
Stockton, CA and DC. The design goal in separation of SprintLink and
ICM was the assurance of symmetry of routing.
[Context?] preservation of next hop in routing is critical feature
(could not wait for IDRP)
Q: What about fears of an ASpath explosion in routing tables? Sean:
Too late. Already here!
More information about the NANOG