TERENA/DANTE TASK FORCE FOR TESTING ADVANCED NETWORKING TECHNOLOGIES

Minutes of the 6th TF-TANT meeting held on the 25th and 26th of November
1999 at ETSIT, Madrid, Spain.

Kevin Meynell - Issue 2

PRESENT

Name                        Organisation        Country
----                        ------------        -------
Hamad el Allali             U.Twente            Netherlands
Werner Almesberger          EPFL                Switzerland
Miguel Angel Sotos          RedIRIS             Spain
Michael Behringer           Cisco               Spain
Graca Caivallio             Cisco               Spain
Massimo Carboni             GARR                Italy
Valentino Cavalli           TERENA              -
Zlatica Cekro               VUB/ULB             Belgium
Nicole Chiminelli           CSELT               Italy
Tryfon Chiotis              GRNET               Greece
Maribel Cosin               RedIRIS             Spain
Luca Dell'Agnello           INFN/GARR           Italy
Tiziana Ferrari             INFN Bologna        Italy
Avgust Jauk                 ARNES               Slovenia
Joop Joosten                CERN                Switzerland
Cees de Laat                U.Utrecht           The Netherlands
Simon Leinen                SWITCH              Switzerland
Lladislav Lhotka            CESNET              Czech Republic
Pedro Lorga                 FCCN                Portugal
Kevin Meynell (Sec)         TERENA              -
Harald Michl                U.Vienna/ACOnet     Austria
Antonio Pinizzotto          CNR                 Italy
Alex van der Plas           Ericsson Telebit    Denmark
Agnes Pouele                DANTE               -
Herve Prigent               RENATER/CRIHAN      France
Juergen Rauschenbach        DFN                 Germany
Filipe Rebelo               FCCN                Portugal
Esther Robles               RedIRIS             Spain
Roberto Sabatino (Chair)    DANTE               -
Rina Samani                 UKERNA              United Kingdom
Robert Stoy                 RUS/DFN             Germany
Celestino Tomas             RedIRIS             Spain
Jean-Marc Uze               RENATER             France
Bert Wijnen                 IBM                 The Netherlands
Wilfried Woeber             ACOnet              Austria

Apologies were received from:

Leon Gommans                U.Utrecht/Cabletron The Netherlands
Dimitrios Kalogeras         GRNET               Greece
Olav Kvittem                Uninett             Norway
Jan Novak                   DANTE               -
Simon Nybroe                Ericsson Telebit    Denmark
Victor Reijs                SURFnet             The Netherlands


1.  APPROVAL OF MINUTES

     The minutes of the TF-TANT meeting held on the 30th of September
     1999 were approved.


2.  STATUS OF QUANTUM & TEN-155

     Roberto reported the connections from Belgium to the Netherlands and
     France had been upgraded to 45 Mbps. The links were operational, but
     the TEN-155 router was not yet installed. Belnet was multihomed to
     TEN-155, and routing was configured to use both circuits. The
     triangle between Spain-France-Switzerland-Spain was due to be
     upgraded to 45 Mbps, whilst an additional 155 Mbps circuit would be
     installed between Italy and Switzerland. A 60 Mbps ATM service was
     being considered between the London PoP and Greece.

     Cees had recently discovered some serious problems with the Germany
     to Netherlands connection. It was only possible to transmit a few
     Kbps during peak hours, and a few Mbps during off-peak hours. It
     transpired that the Ascend switches had been badly configured by KPN
     and were discarding traffic unnecessarily. A temporary workaround
     had been implemented and it was now possible to transit up to 50
     Mbps during off-peak hours. Peak hour performance had improved
     slightly to a few megabits per second, but further investigation was
     clearly required.

     A number of connections were also showing CRC errors. These had been
     traced to certain port adapters on the Cisco routers, and seemed to
     be resolved by upgrading these to Revision 1.

     Wilfried asked whether the cell discard problem was due to the
     switch architecture or the configuration. Roberto replied it was
     more likely to be the latter.

     Wilfried then asked whether the configuration details could be made
     available to the group. Roberto thought this would be difficult, but
     he would talk to him off-line about this.

     ACTION 6.1 - Roberto Sabatino

     Howard reported that due to a price review and as compensation for
     delayed and/or badly performing circuits, a number of connections
     were being substantially upgraded. The ring between UK-Netherlands-
     Germany-France-UK would be upgraded to STM-4. Belgium would also be
     incorporated into the ring as it already traverses Brussels. In
     order to improve resilience, an additional 155 Mbps connection to
     Austria, and an additional 45 Mbps to Hungary would be installed.
     Existing connections to Italy, Sweden and Switzerland would be
     duplicated.

     Even with these upgrades, the total cost of TEN-155 would be 10%
     less than projected. The European Commission had agreed the
     remaining money could be used to fund QUANTUM until October 2000, in
     order to cover the period between the end of the Fourth Framework,
     and the start of the Fifth Framework Programmes.

     Around 160 million euros had been allocated in the Fifth Framework
     for research networking. This was split into RN1 - for production
     networks, and RN2 - for testbeds. The European Commission was having
     difficulty deciding how to allocate this money, as it was felt
     research networking was a special case (i.e. DANTE represents all
     EU/EEA NRNs), but existing regulations required an open call for
     proposals.

     As a consequence, an independent Requirements Advisory Committee
     (RAG) had been established by the European Commission. As the RN1
     funding was clearly intended for DANTE, the Commission officials
     hoped to get a decision from the Commissioners that NRNs should not
     have to bid for their own money. Instead, the RAG would produce a
     specification for the network, and would issue the Call for Tender.
     The membership of about 12 members included David Williams (CERN &
     TERENA), Tomaz Kalin (ARNES), Enzo Valente (INFN), Peter Villemoes
     (NORDUnet), Martin Wilhelm (DFN) and 2 representatives from ETNO
     (European Public Telecommunications Network Operators Association).

     Cees asked whether the successor to TEN-155 actually required EU
     funding as they only covered about 30% of the costs. The commercial
     market moved quickly, and it may be possible to obtain better deals
     if NRNs did not have to involve themselves with the European
     Commission.

     Howard replied the shortfall in funding would not be a problem to
     some NRNs, but others only received funding from their national
     governments if they were part of a EU project. Nevertheless, the
     NRNs were considering this option.


3.  REVIEW OF EXPERIMENTS

     3.1 Differentiated Services

     Antonio presented the interim results of the DiffServ experiment
     (URL?). Tiziana added there had been problems with the SmartBits
     equipment, so they had tried to synchronise their test workstations
     using NTP. This had not proved possible as the workstation clocks
     were not accurate enough. They had therefore attempted to set-up
     connections to and from the same point in order to obtain accurate
     measurements. Unfortunately, this configuration had taken some time
     to debug as there were also problems with the Cisco and IBM routers,
     in addition to the underlying ATM infrastructure.

     Tiziana also outlined the next phases of the experiment. The aim was
     to implement a pre-production service, and validate whether it could
     deliver bandwidth according to specified parameters. The experiment
     should now be extended beyond the initial three sites, with
     additional DiffServ implementations from other vendors (e.g. Nortel
     and Torrent).

     Werner was interested in testing his implementation of DiffServ for
     Linux. He agreed to produce a plan for testing this on the overlay
     network.

     ACTION 6.2 - Werner Almesberger

     Roberto asked when the SmartBits equipment had to be returned.
     Tiziana was not sure, but she would contact Netcom to find out.

     ACTION 6.3 - Tiziana Ferrari

     Tiziana mentioned that a DiffServ BoF had been held during IETF46
     where the results of the TF-TANT experiment had been presented. As a
     result, Rob Nielsen of BCIT and Joe Mambretti of iCAIR had expressed
     some interest in working with TF-TANT. Such collaboration might be
     useful as BCIT were working on a bandwidth broker, whilst iCAIR were
     developing digital video services.

     3.2 Policy Control

     Leon Gommans was not present at the meeting, but Cees mentioned he
     had been participating in the AAA Policy Working Group at the IETF.
     Some work had also been conducted with IP Highway, but this was
     subject to a non-disclosure agreement. Nevertheless, it might be
     possible to establish a private sub-group if TF-TANT members were
     interested in testing the IP Highway implementation.

     3.3 IP over ATM

     Roberto reported the previous two months had been spent fixing
     performance problems on TEN-155. The buffers on the Ascend switches
     had been misconfigured, with the default value of 80 cells proving
     insufficient. This value had been increased to 500 cells, and
     throughput had improved. It was hoped all the problems would be
     fixed by the end of the year.

     Kevin thought it was unsatisfactory that KPN did not know how to
     configure their own equipment properly. TEN-155 would almost be
     finished by the configuration had been sorted out, and if this was
     likely to happen again in the future, DANTE (who had done a good job
     tracing the problems) would be better off managing the equipment
     themselves. Roberto agreed that DANTE could have done a better job,
     but this did not necessarily mean they should do it.

     3.4 MPLS

     Jean-Marc outlined the next phases of this experiment. The first
     activity was to investigate DiffServ mapping on MPLS, followed by
     evaluation of fast restoral. Finally, some interoperability tests
     should be conducted between equipment from Cisco, Nortel Networks
     and Juniper. Ericsson Telebit had offered their laboratory
     facilities for this.

     Unfortunately, Jean-Marc would be leaving RENATER at the end of the
     year, so a new experiment leader would need to be found. In
     addition, the loan period for the Cisco equipment had expired.

     ACTION 6.4 - Roberto Sabatino

     Jean-Marc asked whether the loan period could be extended until
     April 2000. Michael was not sure this would be possible, but he
     agreed to try and arrange an extension.

     ACTION 6.5 - Michael Behringer

     Ladislav mentioned they had tried to run MPLS on their backbone
     network. This had not been successful as a number of routers had
     crashed, and routing information was not advertised outside of the
     MPLS cloud. It had been necessary to remove it after a short period
     of operation.


4.  EXPERIMENT DISCUSSIONS (Parallel Sessions)

     4.1 IP Version 6

     Alex reported that ACOnet had upgraded their router software which
     had created routing instabilities. These problems disappeared after
     downgrading. SWITCH had also experienced routing instabilities, and
     these appeared to be related to certain IOS versions. More
     involvement with Cisco was required to help solve these problems.

     ACOnet, SURFnet and the University of Southampton were participating
     in this experiment using IPv6 over ATM. SWITCH and RedIRIS were
     using IPv4 tunnelling. JOIN (Cisco/Telebit/3Com) would be connecting
     shortly. Nevertheless, the experiment was not progressing as quickly
     as hoped, and it needed more participation.

     In order to progress this experiment, ACOnet volunteered 1.5 FTEs
     for 4 months, DFN said 2 FTEs were being employed for 2 years,
     CESNET thought 2 FTEs could be made available until May 2000,
     RedIRIS volunteered 0.33 of an FTE until May 2000, INFN volunteered
     0.25 of an FTE until May 2000, whilst DANTE thought they could spare
     a few hours per week.

     It was desirable to test as many different IPv6 router
     implementations as possible. Routers from Telebit and Cisco were
     already in use, and 3Com would be shortly. MRTd and Zebra were also
     suggested. It would also be desirable to conduct tests across
     networks other than TEN-155, but one or more NRNs would need to
     provide transit to do this.

     Wilfried thought some further investigation into multihoming
     problems was necessary. These were extremely difficult to track down
     because of the nature of the 6Bone, but he was looking for
     volunteers to help him with this. A request would be sent to the
     mailing list.

     ACTION 6.6 - Wilfried Woeber

     Wilfried also reported he had been unable to get DNSv6 to use
     anything other than IPv4 for transport. There were also other issues
     to resolve before DNSv6 could be used in a production environment,
     and he would summarise these on the mailing list.

     ACTION 6.7 - Wilfried Woeber

     Roberto asked whether a connection to the 6REN should be
     established. Kevin however, thought it was necessary to peer with
     the 6REN at the 6TAP (located at STARTAP), to which TEN-155 did not
     have a connection.

     Lladislav asked whether IPSEC existed for IPv6. Alex replied this
     would be supported by Telebit by the end of Q1. Juergen added that
     some IPv6 implementations for the PC already supported this.

     Juergen mentioned he was compiling a list of IPv6 applications, with
     a view to running them between participants. He asked the group for
     input.

     4.2 Flow-based Monitoring

     Simon said there were some problems with some flow analysis tools.
     The type of data that could be extracted with cflowd was limited as
     it was only possible to produce two-dimensional tables. It was also
     difficult to measure traffic that used non-specified port numbers.
     Furthermore, the high-end Cisco routers only supported aggregated
     NetFlow statistics.

     Roberto thought it might be worth investigating OCXmon. He asked
     whether anyone had any experience of this.

     Simon asked who was interested in using accounting for volume-based
     charging. Avgust replied ARNES were considering this.

     Roberto said that DANTE had to produce statistics for the four AUCS
     interconnects to ensure load balancing was working properly. This
     was being carried out by David Harmelin, but he would welcome input.

     4.3 Multicasting (notes from Valentino Cavalli)

     Robert presented a multicast route monitoring architecture. This was
     currently being implemented by Cisco, and was based on a Manager-
     Agent scheme. Agents monitoring multicast sessions ran at the edge
     of the network, and reported to the Manager. The software for
     Solaris 7.5 was available from http://imj.ucsb.edu/mrm/, whilst the
     protocol specification was available from http://www.ietf.org/
     internet-drafts/draft-ietf-mboned-mrm-00.txt

     Robert proposed to test this architecture on TEN-155, and asked
     which NRNs would be interested in participating. CESNET, INFN,
     RENATER, RedIRIS volunteered for this, while Agnes would check to
     see whether DANTE could provide a workstation in London. DANTE were
     also asked to establish a mailing list for this activity.

     ACTION 6.8 - DANTE

     The Point-to-Multipoint SVC tests were discussed. Such SVC tests
     should ideally be initiated between the PoPs, but the Ascend
     switches currently only supported one SVC per port. It was therefore
     proposed to initiate them on local switches at the NRNs, and use
     tunnelling over TEN-155.

     ACTION 6.9 - Robert Stoy

     Robert also discussed the development of BGMP/MASC. The latest
     implementations were available from http://www.gated.org/ and
     http://netweb.usc.edu/masc/mascd/. The IETF draft specification was
     available from: http://netweb.usc.edu/masc/draft-ietf-malloc-masc-
     04.txt

     Robert would download and compile the latest version of BMGP/MASC,
     and produce a test description.

     ACTION 6.10 - Robert Stoy


5.  DATE OF NEXT MEETING

     The next meeting will be held on the 3rd and 4th of February 2000.
     The venue would probably be CERN, Switzerland.


6.  ANY OTHER BUSINESS

     Cees mentioned that one of their professors had recently won the
     Nobel Prize for Physics. They would therefore like to take a high-
     quality video stream from the awards ceremony in Stockholm on
     Wednesday, 8 December. He asked whether it was possible to obtain
     guaranteed bandwidth from TEN-155 for this.

     Howard replied bandwidth could be reserved via the MBS procedure,
     and Cees should contact Telia. This could usually be set-up very
     quickly, but the connection from the TEN-155 PoP to the Nobel
     Institute would also need to be investigated.

     Roberto reported that as Kevin was leaving TERENA, Valentino Cavalli
     would take-over as Secretary of TF-TANT.

     Roberto also thanked Celes, Esther and RedIRIS for hosting the
     meeting.


7.  ACTIONS FROM LAST MEETING

     5.1  DANTE to provide input to IPv6 report.
          - Done.

     5.2  DiffServ participants to provide input to Internet draft by
          22/10/99.
          - Done.

     4.3  All IPv6 experiment participants to supply information about
          their available equipment, bandwidth and manpower.
          - Ongoing.

     3.8  Victor Reijs to draft document expressing the concerns of the
          research community about STM-4c.
          - Status unknown. Victor was not at the meeting.


OPEN ACTIONS

     6.1  Roberto Sabatino to discuss the TEN-155 switch configuration
          with Wilfried Woeber.

     6.2  Werner Almesberger to produce a plan for testing DiffServ for
          Linux on the overlay network.

     6.3  Tiziana Ferrari to contact Netcom to find out when the
          SmartBits equipment should be returned.

     6.4  Roberto Sabatino to find a new leader for the MPLS experiment.

     6.5  Michael Behringer to try and arrange an extension for the Cisco
          equipment loan.

     6.6  Wilfried Woeber to ask on the mailing list for assistance with
          IPv6 multihoming issues.

     6.7  Wilfried Woeber to summarise DNSv6 issues on the mailing list.

     6.8  DANTE to establish a mailing list for the multicast experiment.

     6.9  Robert Stoy to produce proposal for tunnelling point-to-
          multipoint SVCs over TEN-155.

     6.10 Robert Stoy to produce test description for BMGP/MASC.

     4.3  All IPv6 experiment participants to supply information about
          their available equipment, bandwidth and manpower.

     3.8  Victor Reijs to draft document expressing the concerns of the
          research community about STM-4c.