The specific work to be performed under the NLANR
agreement includes (but is not limited to):
Technical and engineering support and overall coordination of the vBNS
connections at the five supercomputing centers and selected Research and
Education sites.
Testing and measurement of the vBNS performance characteristics which
have
been agreed upon jointly by the NSFNet Program Official and the vBNS
Awardee.
Coordination and oversight of the use of the vBNS as a shared facility
among the supercomputing centers, selected Research and Education sites
and the Awardee. This use as a shared facility should not conflict with
its primary intended use as a research platform.
Coordination and scheduling of utilization of the vBNS by researchers
identified and referred to the Awardee.
Participation in the Research Allocation Committee (VTCC) for the
vBNS.
Coordination of activities at the Supercomputing Centers and selected
Research and Education sites related to the enforcement of the vBNS
Acceptable Use Policy and dissemination of related information.
In November, NLANR participating institutions met during Supercomputing '96
(SC ' 96) and
agreed that future NLANR activities would be focused into three topical
areas: User Services,
Engineering Services, and Research. The FY1997 Program Plan
describing planned activities in these areas is available at http://www.nlanr.net/Reports
/progplan.html.
Where
appropriate, the site specific discussions in this report are divided into
these three categories.
During the first quarter of FY1997, the primary activity involving NLANR
institutions was the
preparation for and participation in demonstrations at SC '96. All sites
participated in this
activity. All sites also provided ongoing engineering and general support
related to the vBNS
operations and applications and participated in the NLANR caching project http://www.nlanr.net/Cache. Another
activities in which
most NLANR sites participated was the High Performance Campus Network
(HPCN) / very high
speed Backbone Network Service Workshop at the University of Colorado at
Boulder, October
22-23, 1996, see http://www.colorado.edu/CNS/HPCN/.
In the October-December 1996 timeframe, the vBNS was very active,
particularly during the
week of Supercomputing '96. Details of vBNS Engineering's work and
accomplishments for the
show are outlined in the Engineering Activities section. The show ran
during the week of
November 18-22, 1996 in Pittsburgh, PA.
Several booths at the show presented demos of supercomputing applications
whose data
processing had previously made use of supercomputing connectivity over the
vBNS. One example
was UCAR's series of 3-D demonstrations of extreme weather events.
Other booths made direct use of the vBNS connectivity to the show floor.
Both the NLANR
booth and the vBNS booth showed a real-time traffic visualization of the
vBNS, using active flow feeds multicast from the deployed vBNS OC-3
monitors. The multicast nature of the input data allowed the booths to
show the same demonstration simultaneously. The display software is
Java-capable, and was developed by Jeff Semke and Jamshid Mahdavi of
PSC.
For a detailed summary of the MCI/vBNS activities during this quarter
(including usage
statistics), see http://www.vbns.net.
Engineering highlights are
included below.
October 1996:
vBNS Engineering continued the vBNS site upgrade process in
preparation of the backbone upgrade to OC-12. At NCAR and PSC, the
Lightstream ATM switch was replaced with a Fore ASX1000, and the Cisco
7000 was replaced with a Cisco 7507 router. One OC-3 monitor was
installed and activated at NCAR, and 2 OC-3 monitors at PSC were
installed and activated.
A vBNS workshop was held at the University of Colorado, Boulder,
where vBNS Engineering presented a briefing on the vBNS. A similar
briefing was also given that week at NANOG.
OC-3 monitor output was modified and enhanced along with the web
page interface to this database.
Further details were established for the design and configuration
for vBNS OC-12 trunking over MCI's ATM HyperStream network.
November 1996:
vBNS Engineering spent most of the month preparing for and
supporting SC'96 in Pittsburgh, PA. Engineers spent several weeks
on-site in a variety of roles, including the design, build,
configuration, and troubleshooting of the show network. Wide-area
connectivity to R&D institutions was also provided for the show.
At the show, vBNS sponsored a booth, with several network
demonstrations, including: a real-time traffic visualization of the
vBNS, using active flow feeds multicast from the deployed vBNS OC-3
monitors; a VRML-based visualization of the vBNS that included new
connections for 1997; a TV-over-ATM demonstration that showed local TV
programming from Richardson, Texas, broadcast over MCI's ATM
HyperStream network; Hippi-over-ATM OC-12 between SDSC and a
Hippi-connected SGI workstation in the booth, using new Netstar OC-12
cards.
vBNS Engineering met with representatives of CANARIE (the Canadian
version of the vBNS) to discuss a wide range of issues regarding
CANARIE/vBNS connectivity across the Ameritech NAP.
December 1996
The last two SCCs, CTC and NCSA, were upgraded with Fore switches
and Cisco 7507s. At CTC an OC3MON was installed to finish the SCC
deployment series of those measurement platforms across the vBNS. All
OC3MONs are providing data now.
NASA Internet began BGP peering with the vBNS at the MFS NAP this
month.
During the first quarter of FY1997, Cornell's activities were focused
on:
Preparing for and participating in SC '96; and
Assisting MCI to upgrade vBNS equipment at CTC
User Services/Outreach:
None reported.
Engineering:
CTC collaborated with MCI and NCSA in the demo of RSVP at the NLANR
booth at SuperComputing 96. Using a Cisco 7507 on loan from MCI,
Cornell loaded Cisco's beta RSVP code and worked with MCI and NCSA on
testing and preparing for the demo. Cornell first connected from the
FDDI of the Cisco to a SGI Indigo running IRIX 6.2 with RSVP and to a
FORE 7000 PowerHub to route to dedicated ethernets on two SGI Onyx
machines running 6.2.
The vic implementation was specially modified by
SGI to support RSVP and they also provided a tool, rtap, for making and
modifying reservations.
The demo at the booth used SGI Indy systems at
CTC and NCSA and a Sun at MCI Reston to generate vic streams to a SGI
Impact on the other side of a MCI-supplied Cisco 4700 router in the
NLANR booth. One vic stream utilized a RSVP reservation, and the other
did not. A machine generating traffic from another ethernet in the
booth created congestion on the dedicated ethernet which connected the
4700 to the Impact.
Numerous problems were encountered. First, the
multicast implementation on the PowerHub at Cornell appeared
incompatible with Cisco's multicast implementation. Since there was no
ethernet interface in the 7507 or FDDI on the Indy systems, they were
connected to MCI's ethernet on the production vBNS router. Since the
production vBNS router at CTC was not RSVP capable, the CTC Indy
systems was used to multicast to the Cisco 7507 at PSC. With non-RSVP
routers in the path, participants found that they could not Guarantee
Bandwidth, and resorted to a Controlled Load. Researchers succeeded in
sending two streams of video from a MCI Sun at Reston to the Impact in
the NLANR booth. Although statistics from the 4700 showed that the
unprotected stream was dropping ten times more packets than the
protected stream, the quality of the video in the two vic windows was
very similar.
During the first quarter of FY1997, Cornell's activities were focused
on:
Supporting NCAR applications and initiating discussions regarding a
gigaPoP at NCAR/UCAR; and
Supporting various equipment and
routing upgrades to the vBNS.
User Services/Outreach:
Chris Fair and Paul Hyder met with David Wood, Caren Litvanyi
(University of Colorado), and Chris Garner and Doug Kerry (SuperNet,
Inc.), to discuss full route peering for the University of Colorado,
Boulder through NCAR to the vBNS. The decision was made to recommend
that MCI configure the University of Colorado to peer directly with the
vBNS.
Chris Fair, Marla Meehl, and Paul Hyder met with Ken Klingenstein and
representatives of the University of Wyoming, Colorado State
University, Denver University, and SuperNet, Inc. to discuss
involvement in the construction of a GIGAPOP architecture at
NCAR/UCAR.
Chris Fair also presented issues relevant to High Performance Campus
Connections to broadband, high bandwidth-delay product networks at the
HPCN/vBNS Workshop (see above)
Internet Data Distribution (IDD) Project - Chris Fair and
Marla Meehl met with UNIDATA personnel to discuss the resubmission of
application for vBNS experiments involving the Unidata Internet Data
Distribution (IDD) System. This is a result of the announcement of the
NSF awards to university sites under the NSF Connections '96 program
served by the IDD system. The URL for the IDD homepage is
http://www.unidata.ucar.edu/projects/idd/index.html. The URL for
the IDD vBNS Friendly User application is http://www.unidata.u
car.edu/projects/i dd/vbns.html.
Distributed Mesoscale Prediction (DMP) Project - The Major
Research Instrumentation (MRI) Program is a National Science Foundation
program that "assists in the acquisition or development of major
research instrumentation at U.S. institutions". DMP is a project to
obtain MRI Program funding to build a Distributed Numerical Weather
Prediction Laboratory (DNWPL) that uses the Penn State/NCAR MM5 model
running on dedicated computers initially distributed at NCAR, the
University of Arizona, the University of Utah, and the Desert Research
Institution. This project assumes that the participating institutions
will be connected to the vBNS and that the vBNS is the data
transportation vehicle for effecting data transfer among the
distributed model instances. Data flow would be radial, with NCAR at
the center. The bulk of data would flow from the other three sites to
NCAR, with a smaller amount of data flowingaway from NCAR. The
aggregate of data flow at NCAR is estimated to be no more than 55 Mbps.
Engineering:
Chris Fair's efforts during this quarter included:
working with MCI Engineering in replacement of vBNS Fore Systems
ASX-200 ATM switch with Fore Systems ASX-1000 ATM switch and Cisco Systems
7000 router with Cisco Systems 7507 router.
working with MCI Engineering on installation and configuration of
vBNS OC-3 monitor at NCAR.
working with Laura Cunningham to provide ATM through path VPI 50 on
NCAR ATM switch connected to vBNS test3 for vBNS support to the University of
Colorado. CU and MCI can now configure VCIs as necessary without specific
NCAR intervention. This will also facilitate ease of configuration for BGP
peering relationship between University of Colorado and MCI.
beginning performance measurement and analysis between PSC, NCAR and
SDSC to benchmark specific bandwidth performance, availability and
utilization toward vBNS performance goals.
identifying and correcting the power supply failure of vBNS Fore Systems
ASX-1000 ATM switch. Worked with MCI Engineering to install second AC power
supply in NCAR Fore Systems ASX-1000 ATM switch.
NCAR/UCAR personnel continue to implement the NCAR Network Re-Engineering
Plan and have replaced aging Cisco Systems AGS+ multiprotocol routers with new
Cisco Systems 7507 series routers. This is in conjunction with the deployment
of Cisco Catalyst 5000 Ethernet switches and Cisco LightStream 1010 ATM
switches. 10 and 100 Mbps connections are being installed to the desktop and
servers. The Ethernet switches are temporarily being ATM trunked to
NCAR/UCAR's Fore Systems ASX-200. All deployment of Catalyst 5000 Ethernet
switches will ATM trunk to new LightStream 1010 ATM switches using LAN
Emulation. This redeployment of the NCAR/UCAR campus network infrastructure
will provide state of the art, high performance connectivity to present and
future NCAR/UCAR vBNS applications.
Research:
None reported.
National Center for Supercomputing Applications
(NCSA)
During the first quarter of FY1997, NCSA's activities were focused on:
Preparing for and participating in SC '96;
Facilitating activities relating to international
connections;
Configuring an OC3mon monitor, including initiating efforts to port it
to Linux
User Services/Outreach:
NCSA staff participated in the following NLANR related efforts in
preparation for, and during,
Supercomputing '96:
CEWES/NCSA - Work is underway at CEWES (the Army Corps of
Engineers
Waterways Experiment Station in Vicksburg Mississippi) to develop a
three-dimensional
simulation model to address eutrophication issues (high concentration of
nutrients in the deeper
layers) in the Chesapeak Bay. At SC96 ImmersaDesk systems from CEWES and
NCSA were
transported to the show floor and linked together via IP Multicast for a
collaborative tour of the
Chesapeake Bay. This was a precursor to a live demonstration planned for
the end of February
when the collaboration will take place between CAVE systems at the two
sites over the vBNS
and IDREN networks. NCSA NLANR staff worked together with the applications
developers in
the design, testing and support of the network specifics of the
collaborative application.
NLANR Booth - NCSA aided in the planning, setup, tear down and
manning of the
NLANR booth.
CANARIE Demonstration - NCSA staff worked with Michel Savoie of
the
Communications Research Centre, Ottawa, Canada (CRC Badlabs)
http://www.crc.doc.ca on a
planned demonstration of the CANARIE link to the vBNS. The demonstration
was to feature the
ISABEL collaborative tool but unfortunately the link was not activated in
time. Still testing and
preparation time was spent prior to the SC96 show.
RSVP Testing - NCSA participated in a RSVP testbed over the
vBNS for Supercomputing '96 with MCI's vBNS Engineering group and CTC.
The testbed involved installation of beta software on host at NCSA,
CTC, MCI and the SC96 show floor as well as installation of RSVP
capable Cisco routers at all four sites. Participants were able to
demonstrate basic interoperability of the different RSVP
implementations and test operation of RSVP over a non-RSVP aware
wide-area network (the vBNS). Tests with sending RSVP versus non-RSVP
video onto the show flow showed that the resulting improvement was
measurable but not-significant. NCSA biggest contribution was with the
pre-show testing between NCSA and MCI. NCSA led the testing of both
the Silicon Graphics IRIX and Sun Solaris host RSVP code.
In support of both the GIBN initiative and the vBNS programs, NCSA
spent time first advocating the interconnection of the vBNS and
CANARIE networks and then later on assisted with the connection details
and politics. The link was activated after the SC'96 show.
Researchers at NCSA are already working with colleagues at Canadian
research sites on network based collaborations and expect the first use
of the link shortly. Assistance is also being provided to researchers
including Rod Whiting from the University of Calgary on a collaborative
medical application.
NCSA organized a joint meeting between the CANARIE and vBNS technical
committees at
SC'96. The immediate results of the meeting were a better understanding of
the common issues
both groups face and a strong interest in a closer working relationship.
Engineering:
In addition to the engineering required for SC'96, NCSA was active in
the development of a Route Server Tool Set. This tool set displays
routes the vBNS is handing customers, a very useful tool for the
administration of vBNS routes. A web server at NCSA displays routes
supplied by the vBNS. It additionally allows users to perform
traceroute to an arbitrary site in order to determine whether or not
the site is reachable via the vBNS. (see http://rs.ncsa.uiuc.edu) NCSA will
continue to add additional functionality to this simple operational
tool as the need arises.
NCSA also helped establish the CANARIE link into MREN/AADS (see above).
Research:
OC3Mon - Jon Dugan (NCSA) put together a PC, modifying the software to
allow it to run LINUX OS complete with FDDI and ATM interfaces. The
porting of the OC3mon and Network Flows Software to this LINUX system
was initiated during this period.
During the first quarter of FY1997, PSC's activities were focused on:
Organizing the NLANR booth at SC '96.
Upgrading the vBNS Mbone infrastructure to PIM in advance of
SC'96
User Services/Outreach:
The SC'96 activities described below were in support of NLANR's overall
outreach to the
computing/networking community. In addition to these efforts, Jamshid
Mahdavi participated in
the HPCN Workshop in Boulder, October 22-23, 1996 (see above).
Engineering:
During this period, PSC's NLANR efforts focused heavily on Specific
demonstrations in which
PSC personnel participated included:
A Harvest WWW cache on the show floor, available for use by show
participants.
A realtime OC3mon demo, showing live flow data collected from the
OC3mon and multicast
out to the show floor. Because of the multicast nature of this
application, researchers were able
to show
live flow data in both the NLANR and MCI booths. With the live OC3mon,
NLANR was able to
track high performance applications utilizing the vBNS in real time
throughout the show.
A RSVP QoS demo, attempting to establish a resource reservation
through the vBNS test
network to Reston, VA. This (apparently simple) attempt to use QoS
revealed a number of
problems in the current implementations of QoS and RSVP. Although MCI was
finally able to
establish an end-to-end reservation, the effort highlighted areas where
more work is needed. MCI
continues to pursue these areas.
A live distributed application demo between several booths at the
show (NLANR, Sandia)
and the University of Stutgart. This demo used COVISE, a collaborative
engineering tool, to
allow researchers at the various sites to simultaneously manipulate a model
of an automobile
engine. Unfortunately, a high-speed network was not available between the
show floor and the
site in Germany, greatly limitting what could be accomplished.
PSC attempted to bring up an IP v.6 node on the show floor, but it was
not possible to
extend the 6bone through the vBNS to the floor with the amount of time
available.
However, participants hope to see some IP v.6 infrastructure in the vBNS
soon. (Note: In
January, MCI
placed an order for 6bone tunnel machines for each vBNS site).
In conjunction with a PSC demo, participants also attempted to setup
end-to-end PVCs
through the Sprint ATM network to Sandia and Oak Ridge National Labs. This
turned out to be
a
cumbersome and complicated effort and underlined the non-scalability of
current ATM
technology and the lack of debugging tools.
PSC also continued to maintain the TCP performance tuning WWW site, which
has become quite
popular (and has apparently been published on CDROM in India).
During the first quarter of FY1997, UCSD's activities were focused on:
Preparing for and participating in SC '96 and
Presentations at the FNCAC, NANOG, and IETF
Chairing a BOF at the IETF
Supplemental Proposal to NSF; Hiring of new personnel starting 2Q97
User Services/Outreach:
UCSD/NLANR personnel delivered presentations on measurement at the
following events during
the first quarter of FY1997:
Federal Networking Council Advisory Committee (FNCAC) on October 20-21,
1996, see
http://www.fnc.gov/FNCAC.html
North American Network Operators Group (NANOG) on October 24-25, 1997, see
http://www.nanog.org
Coordinating Committee for Intercontinental Research Networking (CCIRN) on December
12, 1997
Duane Wessels participated in an international Web Caching Workshop in
Warsaw, Poland
in October 1996, sponsored by the Interdisciplinary Centre for Mathematical
and Computational
Modelling in Warsaw. Attendees were primarily from European countries.
About half of the
workshop consisted of talks relating to works-in-progress or future
protocols (HTTP/1.1,
HTTP-NG). The other half consisted of reports from various members
describing their particular
cache configurations. Duane delivered four presentations: ICP, Squid,
Neighbor selection
algorithms, and the NLANR Caches.
In November, Duane Wessels participated in Supercomputing '96 in
Pittsburgh, PA and was
instrumental in demonstrating a Web cache for show participants. He also
presented a real-time
visualization of cache traffic flows at the booth. This demo used
geomview to display
requests between cache peers around the globe.
Duane Wessels and k claffy also co-chaired the (see below).Duane Wessels
and k claffy attended
Internet Engineering Task Force (IETF) #37 in San Jose, CA in December and
co-chaired the
Birds of a Feather (BOF) session on ICP (see below). Duane also attended
the HTTP working
group meetings and met with several Squid users and other software
developers. Duane also
delivered a presentation on caching to the caching working group of the
CCIRN.
Engineering:
Focused on preparing for web caching demonstrations at SC'96 (see above).
Rich Gallup extended SDSC route server for use with vBNS peering.
Research:
Measurements - During this quarter we continued
collaborating with MCI/vBNS on OC3mon, BBN/ANS on cflowd, and
Merit on routing tools and data visualization. Continued work
on IPNmoo. Began porting Oc3mon to netbsd.
Initiated discussions with a private company (InterVU) regarding
collaborating on enhancements to its internal video
distribution performance measurement tool to make it more
robust and generalizable to other measurement/visualization
requirements.
Set up NLANR-NorthernCalifornia site, also home of Caida,
with connectivity to Palo Alto Internet Exchange.
We are peering (receiving routes only) with routers at
Digital, Palo Alto and investigating visualization
techniques for the data.
Cache Hierarchy Status - The use of the NLANR caches continues to
increase. We
have approximately 100 participant sites using our caches. All together,
they process one million
requests per day and serve 12 Gbytes.
In July 1996, UCSD/NLANR started operating a cache on a machine owned
by Carl Malamud, as a part of the Internet 1996 World's Exposition.
This machine was located one hop away from MAE-East and quickly became
very popular with European sites. Unfortunately, as the Exposition
ended, the machine was moved and were no longer able to use it for a
cache node. That cache received the most traffic (per day) of any
caches -- a clear indication of the importance of locating caches near
exchange points.
Squid Caching Software - Squid version 1.1 was in a beta
testing phase during October - November, 1996. In December, the
official Squid version 1.1 was released. This version included the
following important features:
Improved handling of stale data. Squid-1.0 always removed
stale objects from the cache. Squid-1.1 keeps the stale
objects and issues If-Modified-Since requests for them
if they are requested again.
Numerous cache sites wanted the ability to treat a peer cache
as a parent for some requests and as a sibling for other
requests. Parents can transit a request on behalf of a peer,
siblings cannot. Consider a pair of caches, one in Germany
and another in Poland. The German cache would like to be a
parent for URLs in Germany, but a sibling for non-German URLs.
With Squid-1.0 a peer was always a sibling or always a parent.
Similar to the feature describe above, cache administrators
wanted to be able to enforce the parent/sibling relationship
on their peers. The existing access controls allowed them
to specify which peers were allowed access, and for which
sort of requests. But it was not possible to allow only
requests for *cached* objects and deny reqeusts which
would require transit.
Added a hook to rewrite requested URLs. There were numerous
requests for the ability to rewrite a requested URL. This
was implemented as an external process so that it could be
very customizable on a per-site basis. One application of this
is filtering (or censorship). The redirector program can be
written to recognize certain URLs and redirect the request
to a local Web page describing why access has been denied.
Another interesting application is related to mirroring highly
popular data. Often, highly popular URLs are not always found
in a cache because there are many different mirror sites, each
with a different URL. For example,
are all URLs which point to the same file. The redirector
might look at only the filename and return a URL to a single
local mirror copy. This approach is a bit controversial, and
requires a lot of manual configuration.
Improved access control features. This version performs
ident (RFC 931) lookups, as well as reverse IP Lookups.
Changed disk directory structure. Mark Treacy mark@aone.com.au did some simulations
and found that
a
two-level directory structure performed considerably better
than the single-level approach in Squid-1.0.
Squid uses an IP-cache to avoid the need for frequent DNS
lookups. This bothers some people because the DNS has native
caching features which Squid does not take advantage of.
Specifically they worry that Squid will cache an address longer
than the TTL value returned with the DNS lookup. Normally
Squid does not have access to the DNS TTL value because the
gethostbyname() library function does not provide it. However,
Endre Balint Nagy bne@CareNet.hu came up with
a relatively
simple modification to the BIND code which allows this. If
cache sites wish to apply a simple patch to BIND then their
cache can use true DNS TTLs.
When a DNS lookup results in numerous IP addresses, Squid now
cycles through all addresses in a round-robin fashion.
We have received a lot of assistance from the user community.
The following names have contributed time and effort to
the project:
Henrik Nordstrom, henrik.nordstrom@ida.his.se
Mark Treacy, mark@aone.com.au
Daniel O'Callaghan, danny@miriworld.its.unimelb.EDU.AU
George Michaelson, ggm@connect.com.au
Taavi Talvik, taavi@uninet.ee
Luke Howard, lukeh@vurt.schnet.edu.au
Assar Westerlund, assar@pdc.kth.se
James R Grinter, jrg@demon.net
Shigechika Aikawa, shige@luck.imasy.or.jp
Russell Vincent, vincent@ucthpx.uct.ac.za
Edward Moy, moy@parc.xerox.com
Jean-Francois Micouleau, Jean-Francois.Micouleau@utc.fr
Ansgar Hockmann, Ansgar.Hockmann@hrz.uni-dortmund.de
Joe Ramey, ramey@jello.csc.ti.com
Kolics Bertold, bertold@tohotom.vein.hu
Doug Urner, dlu@bsdi.com
Markus Gyger, mgyger@itr.ch
Richard Huveneers, Richard.Huveneers@hekkihek.hacom.nl
Markus Stumpf, maex@Space.NET
Wojtek Sylwestrzak, W.Sylwestrzak@icm.edu.pl
Brian Denehy, B-Denehy@adfa.oz.au
David Luyer, luyer@ucs.uwa.edu.au
Carson Gaspar (carson@lehman.com, carson@cs.columbia.edu)
Pete Bentley, pete@demon.net
Alexander B. Demenshin, aldem@barnet.kharkov.ua
Tony Lorimer, tlorimer@au.mdis.com
Igor Vinokurov, igor@cs.ibank.ru
Russell Street, r.street@auckland.ac.nz
Cord Beermann, cord@cc.fh-lippe.de
Stephen R. van den Berg, srb@cuci.nl
Jon Thackray, jrmt@uk.gdscorp.com
John Saunders, johns@rd.scitec.com.au
Endre Balint Nagy, bne@CareNet.hu
Geoff Keating, Geoff.Keating@anu.edu.au
Arjan de Vet, Arjan.deVet@adv.IAEhv.nl
Mike Groeneweg, mikeg@scorpion.murdoch.edu.au
Michael Mansour, mic@shell.gnxs.com.au
Ralph Loader, loader@maths.ox.ac.uk
Miguel A.L. Paraz, map@iphil.net
Jonathan Larmour, JLarmour@origin-at.co.uk
Andres Kroonmaa, andre@ml.ee
Stewart Forster, slf@connect.com.au
Anthony Baxter, arb@connect.com.au
Martin Hamilton, martin@mrrl.lut.ac.uk
Ed Knowles, ed@fatboy.geog.unsw.edu.au
Neil Murray, neil@aone.com.au
Jonathan Larmour, JLarmour@origin-at.co.uk
Andreas Lamprecht, Andreas.Lamprecht@siemens.at
Ron Gomes, rrg@ny.ubs.com
Jan Niehusmann, jan@anduin.gondor.mcs.de
Taketo Kabe, kabe@shiratori.riec.tohoku.ac.jp
Mukaigawa Shin'ichi, shin@nff.ncl.omron.co.jp
Lutz Donnerhacke, lutz@iks-jena.de
ICP - An Internet Draft on the Internet Cache Protocol (ICP) was
written and
submitted in November 1996. This draft was intended to document the
current ICP
implementation and usage. It was not intended as a proposal for future
evolution of the protocol.
The Internet Draft was discussed during a BOF at the IETF in San Jose, CA
in December.
Originally, the BOF was intended as an informal meeting of parties who
planned to utilize ICP in
a cache product. The meeting was intended todiscuss what would be an
appropriate forum (if
any) to advance the development of ICP as a protocol. The BOF was actually
attended by over
120 people, many of whom assumed that since we were already at the IETF, we
had choosen the
IETF as the forum for standardizing ICP. The outcome of the BOF was
concensus that an ICP
working group should be formed and ICP would be written up as an
informational RFC under the
Operations area.
As of December 1996, the NLANR cooperative agreement had not yet been
transferred from
General Atomics to the University of California, San Diego. We anticipated
that this transfer will
take place in the second quarter of FY1997. At the conclusion of that
quarter, there will be a
summary of resources expended to date for CTC, NCSA, PSC, and UCSD.
NCAR is managed under a separate grant. As of this report, it had not yet
received its funding for
FY1997.
Mike Berry, AT&T Canada, mberry@attcanada.com
Randy Butler, NCSA, rbutler@ncsa.edu
Bilal Chinoy, SDSC, bac@sdsc.edu
Anne Hammond, University of Colorado, hammond@colorado.edu
Doug Hughes, Canarie, Inc., dihughes@canarie.ca
Tracy Hughes, MCI, thughes@mci.net
Wendy Huntoon, PSC, huntoon@psc.edu
John Jamison, MCI, jjamison@mci.net
Bruce Johnson, CTC, bbj@tc.cornell.edu
Charles M. Loo, MCI, 5656578@mc.mail.com
Steven Ladouceur, Bell Canada, sladouceur@on.bell.ca
Mark Luker, NSF, mluker@nsf.gov
Jamshid Mahdavi, PSC, mahdavi@psc.edu
Matt Mathis, PSC, mathis@psc.edu
Marla Meehl, NCAR, marla@ucar.edu
Greg Miller, MCI, gmiller@mci.net
Don Mitchell, NSF/DNLRI, dmitchel@nsf.gov
Randy Nicklas, MCI, rcn@mci.net
Jeff Semke, PSC, semke@psc.edu
Mike Shaver, Ingania Communications, shaver@ingenia.com
Jerry Sobieski, University of Maryland, jerrys@umiacs.umd.edu
Bill St. Arnaud, Canarie, bill@canarie.ca
Jim Stacey, Acorn, jstacey@morgan.ucs.mun.ca
George Strawn, NSF/DNCRI, gstrawn@nsf.gov
Duane Wessels, UCSD/NLANR, wessels@nlanr.net
Linda Winkler, Argonne, lwinkler@anl.gov
Rod Wittig, WurcNet/Univ. of Calgary, wittig@acs.ucalgary.ca
Steve Wolff, Cisco, swolff@cisco.com
Wolfgang Ziegler, GND ziegler@gnd.de
1) Introduction of attendees.
2) Randy reviewed the agenda for the meeting.
3) Canarie review: Bill Arnaud
- TAD: Technology Application Development funded
- Network Activities
- CAnet (like NSFnet) - dissolved 4/1/97
- Canarie high-speed testnet - like vBNS; connect regionals not
individual sites
- Bell Canada provides major backbone
- International link to Europe (Teleglobe)
- Link to US Ameritech NAP is expected - maybe link to Spring NAP
- CAnet II is being planned like Internet II in US for broader
connectivity - DS3 to OC3 to OC12
- Gigapop detail concept described
4) Questions of Canarie
- How to get to Europe via Canarie? -- Randy - offer to transit thru
network
- encourage 3-way projects including Canada
- talk to Steve about getting routing set up
- TNOC application process thru Canada
- scheduled in Europe -- up and down - JAMES network in Europe
has to approve
- not really end-user ready
- G7 applications in but no AUP, single source of application,
network, etc.
5) Mark Luker
- agreement reached to connect Canada and vBNS - open connectivity
- vBNS changes
1. add more connectivity - 5 SCCs to 100 sites
2. network going for OC3 to OC12
4. new stability - production network; stable; leading edge
5. new policies
- AUP
- all vBNS to vBNS traffic authorized partners
- approval process based on who non-attached
sites need to work with
- who attached sites need to talk to,
international or non-connected
sites
- no transit traffic
- immediate needs are not met by this network at
the international level....G7 comes in:
- unfunded mandate at US level
- no carrier donations from US side
- no GIBN network currently there or
planned
- international connections all viewed
as important
6. Interconnect agency networks
- How does int'l connections program fit in ("hopes to fit in" -Strawn)
- NSF looking at domestic and int'l infrastructure together
- how to work with other agencies for approvals, etc.
- work with PIs of current apps to derive procedures
- How has Canada done this? -- Randy
- TNOC like vBNS - schedule/approve projects now
- AT&T and Bell Canada manage network -- end-user
staff (funded to do so)
- small team of staff for funding only at Canarie
6) Rick Wilder -- vBNS scaling up
- 5 nodes at SCCs and 4 nodes at NAPs
- now unknown equipment (wasn't originally)
- less unity, more team work
- MCI gearing up to handle diversity
- more coordination, re-organization
- separate planning and operational committees suggested
- Matt noted that it looks a lot like T1 - paranoid routing
- MCI not planning end-to-end support
- routing policy evolution
- primary institution can send all traffic between themselves
- secondary institution can send only to primaries
- risk increases as more and more sites are added and higher
percentage of traffic and dependence is on the vBNS
- each site can filter routes themselves - need list of nets
to expect from MCI or collect themselves
- routing and security issues - ATM dangerous, difficult
- Can vBNS sites reject data? - yes
- need extended technical group with new connections sites
- core advisory group within this technical group
- Will this include other providers? - yes
6) Connections Sites
- When will they be converted?
- MCI working out contract issues
- University of Pennsylvania first to be connected -- 60 days
- Is there a restriction on how soon sites must connect?
- review process
- 7131 deadline
- panel 2 weeks later
- awards just coming out 11/96
- status of connectivity will be added to each connection site
web page - Chicago in Spring '97
