 |
Minutes of 2nd TF-TANT Meeting |
 |
Message-ID: <4.1.19990321150537.00b9ae40@popper.terena.nl>
Date: Sun, 21 Mar 1999 15:06:29 +0100
From: Kevin Meynell
Subject: Revised minutes of the 2nd TF-TANT meeting
To: TF-TANT@TERENA.NL
TERENA/DANTE TASK FORCE FOR TESTING ADVANCED NETWORKING TECHNOLOGIES
Minutes of the 2nd TF-TANT meeting held on the 25th and 26th of January
1999 at the TERENA Secretariat, Amsterdam, The Netherlands.
Kevin Meynell - Issue 2
PRESENT
Name Organisation Country
---- ------------ -------
Bob Aiken Cisco United States
Pantelis Balaouras GRNET Greece
Michael Behringer Cisco United Kingdom
Rui Bettencourt FCCN Portugal
Mauro Campanella INFN Milano Italy
Zlatica Cekro VUB/ULB Belgium
Howard Davies DANTE -
Rob van Dinther KPN The Netherlands
Armando Domingues FCCN Portugal
Francis Dupont INRIA France
Alain Durand IMAG/G6 France
Tiziana Ferrari INFN Bologna Italy
Dusan Gabrijelcic IJS Slovenia
Silvia Giordano EPFL Switzerland
Leon Gommans U.Utrecht/Cabletron The Netherlands
Christoph Graf (Chair) DANTE -
Jaap Heffels KPN The Netherlands
Dimitrios Kalogeras GRNET Greece
Olav Kvittem Uninett Norway
Cees de Laat U.Utrecht The Netherlands
Ladislav Lhotka CESNET/USB Czech Republic
Olivier Martin CERN Switzerland
Vassilis Merekoulias GRNET Greece
Kevin Meynell (Sec) TERENA -
Jan Novak DANTE -
Simon Nybroe Telebit Denmark
Herman Pals KPN Research The Netherlands
Juergen Rauschenbach DFN Germany
Victor Reijs SURFnet The Netherlands
Esther Robles RedIRIS Spain
Roberto Sabatino DANTE -
Robert Stoy RUS/DFN Germany
David Sutherland KPN The Netherlands
Jean-Marc Uze RENATER France
Bert Wijnen IBM The Netherlands
Wilfried Woeber ACOnet Austria
Apologies were received from:
Brian Carpenter IBM UK United Kingdom
Simon Leinen SWITCH Switzerland
Guenther Schmittner JKU/ACOnet Austria
Istron Tetenyi HUNGARNET Hungary
Bernard Tuy CNRS France
1. APPROVAL OF MINUTES
The minutes of the TF-TANT meeting held on the 5th and 6th of
November 1998 were approved.
2. STATUS OF QUANTUM & TEN-155
Howard reported the QUANTUM contract had been received from the
European Commission in mid-January having been dated the 28 December
1998. As expected, the project start date was specified as the 1st
of October 1998 which meant a number of deliverables were already
due.
The first lines for TEN-155 had already been delivered. These were
the lines in the UK-Netherlands-Germany-France-UK loop, and the
lines between the Netherlands, Switzerland, Germany and Sweden. The
line from TEN-155 to the US was also operational. Lines from the
Netherlands to Belgium, Germany to Sweden, Germany to the Czech
Republic, and Greece to the Netherlands, were due shortly.
Unfortunately, the line between Austria and Germany had failed
testing and was having to be re-tested. This would cause a knock-on
delay for the Austria to Hungary connection. In addition, the
Hermes supplied line between Germany and Italy had not yet been
delivered. As Hermes had also failed to supply the connection from
the Netherlands to Luxembourg on-time, this had raised some concerns
about their performance.
There were still problems with Spain, Portugal and Slovenia that
needed to be resolved before the old TEN-34 contract expired at the
end of March. Spain was due to be connected to TEN-155 sometime in
April, but a contract had not yet been signed with Portugal Telecom
as they were quoting extortionate rates. In Slovenia, the
telecommunications monopoly were offering reasonably acceptable
prices, but they wished to retain ownership and control of all the
equipment. This was mainly because they considered ARNES (the
Slovenian NRN) a threat to their business.
Three more countries were also planning to connect to TEN-155.
Cyprus and Israel would be connected in the context of the EU-funded
Q-MED project, whilst POL-34 (the Polish Supercomputer Network)
would provide the access point in Poland.
On the equipment front, the Ascend ATM switches were showing some
problems under high traffic loadings. This was because each
interface had a single processor queue that was equally balanced
between CBR and UBR, but TEN-155 was currently configured to solely
run UBR. It was likely this problem could be solved by adding more
interfaces. In addition, the new Cisco 12000 router in New York had
a VC queueing problem. If this could not be resolved, a Cisco 7500
would be moved from Washington D.C.
The roll-out of TEN-155 also meant interconnections with some
commercial networks had to be re-negotiated. BT still peered with
TEN-155, but AUCS had stopped routing traffic to Ebone despite
agreeing to continue doing this for another six months.
Unfortunately, commercial networks did not appear to be willing to
sign zero-cost peering agreements any longer.
Wilfried was concerned about when the Cisco 7000 series routers were
due to be replaced on TEN-155, especially given the problems with
the new 12000 series routers. Howard replied there was no defined
timetable, but the 7000s ultimately had to be replaced because they
would not be adequate. They were looking at equipment from other
vendors, but their requirements were quite specialist.
Dimitrios expressed unhappiness at the number of routes to
commercial networks that had been lost. He said some customers had
been charged 10 million drachma because their traffic was having to
use backup routes.
3. MANAGED BANDWIDTH SERVICE
Howard reported the QUANTUM Policy Committee had approved use of the
TEN-155 Managed Bandwidth Service (MBS) for TF-TANT activities. The
amount of bandwidth dedicated to the MBS had still not been defined,
but it would be in the order of 15-20% as originally envisaged. An
alpha test phase would commence in January with the MECCANO Project
as the users, and this would be followed by a beta test phase with 4
or 5 users (possibly including TF-TANT). It was envisaged the full
service would start from May onwards.
In order to capacity plan, some information would be required from
the NRNs on how much bandwidth they planned to dedicate to the MBS.
In addition, it was important to obtain information on local
connectivity. It was no coincidence that the projects given early
access to the MBS had good access to their national PoPs.
Tiziana asked whether a permanent 2 Mbps overlay network could be
established for the TF-TANT activities. Howard thought this would
not be a problem.
Jean-Marc asked whether specific test equipment could be located at
a PoP. Howard replied this was generally not a problem, provided
there was sufficient space to accommodate it.
4. REVIEW OF EXPERIMENT DESCRIPTIONS
4.1 RSVP
Simon Leinen was not present at the meeting, but Christoph reported
there did not appear to be much interest in RSVP as a separate
activity. Simon had proposed to amalgamate this experiment with the
RSVP to ATM SVC mapping experiment.
Victor thought RSVP was experiencing an identity crisis. It did not
appear to scale well and it would be difficult to manage. In
addition, the QBone was currently drawing attention away from it.
Bob said that some people were looking at using RSVP as a signalling
protocol in conjunction with DiffServ. RSVP had problems handling
many small flows, so it may not scale to backbones in general.
Nevertheless, it may be possible to use it in a core network with
small subsets of users, or in an campus environment.
Christoph concluded there was not enough support for this experiment
on its own. Simon should therefore contact Tiziana to determine what
activities could be included in the RSVP to ATM SVC Mapping
Experiment.
ACTION 2.1 - Simon Leinen
4.2 Multicasting (IP and ATM)
Robert presented his proposal for establishing a native multicast
network over TEN-155 (http://www-ks.rus.uni-stuttgart.de/TF-TANT/
Multicast-test-spec.htm). This would allow performance issues and
multicast routing protocols to be investigated.
Jan added the current TEN-155 multicast overlay had to be
re-designed anyway, as the Mbone feed from MCI would shortly
disappear. They had an offer of a DVMRP feed from UUNET, but this
required tunnels to be moved around.
Christoph suggested that Robert and Jan should specify the TEN-155
multicast facilities together.
ACTION 2.2 - Robert Stoy and Jan Novak
4.3 Differentiated Services
Tiziana presented her proposal for testing Differentiated Services
(http://www.cnaf.infn.it/~ferrari/tfng/diffserv.html). She proposed
to investigate IP precedence-based QoS mechanisms and DiffServ
during the 1st Quarter of 1999, and then a mixed DiffServ-intServ
architecture during the 2nd Quarter. There were also some issues
related to policy and security that needed to be investigated.
Christoph said the point of DiffServ was that applications did not
need to know about it. Bob added that policy servers were
responsible for allocation of bandwidth, although these were still
being developed.
Christoph then asked what this experiment required from QUANTUM.
Tiziana replied that a CBR class of service was required, but the
tests would not require equipment to be located at the PoPs. At a
later stage however, it might be useful to establish a domain over
the core network itself.
Mauro asked about the type of applications being considered for the
tests. Tiziana suggested involving the Condor (distributed
computing) and MONARC (simulation and modelling) projects for this.
Bob added it might be interesting to include the Globus Project
(http://www.globus.org/) which had similar aims to the Condor
Project.
Cees de Laat asked to be involved in this experiment as the DYNACORE
Project (real-time control) required QoS for its activities. Tiziana
agreed to integrate the project into her proposal.
ACTION 2.3 - Tiziana Ferrari
Finally, Silvia mentioned that EPFL were developing a DiffServ
implementation. They were interested to become involved in this
experiment.
4.4 RSVP to ATM Mapping
Tiziana presented her proposal that aimed to determine whether the
integration of IP and ATM QoS mechanisms was viable (http://www.
cnaf.infn.it/~ferrari/tfng/rsvp-atm.html). Tunnelling could be used
for this experiment, but per-VC shaping would be necessary over
TEN-155.
Ladislav asked how the RSVP to SVC mapping would be implemented.
Tiziana replied that EPFL were currently developing an RSVP add-on
to the Linux ATM distribution which should be available shortly.
Bob added that Cisco would also be including RSVP mapping facilities
in IOS.
4.5 IP Version 6
Simon presented his proposal for implementing an IPv6 service over
TEN-155 (http://www.dante.net/tf-tant/experiments/IPv6/). This would
be closely mapped to the existing IPv4 service and would consist of
a single routing domain running EBGP.
Alain added the goal of the experiment was to establish a
production-quality network that would allow IPv6-compliant
applications to be tested. The 6Bone currently provided some
facilities, but this was tunnelled over the regular Internet and was
not very robust. Internet2, vBNS, CANARIE, ESnet, CAIRN and WIDE had
therefore formed a native IPv6 network known as the 6REN (http://
www.6ren.net/) which enforced certain conditions. It would be useful
to have something similar in Europe which could be interconnected
with the 6REN via the STARTAP in Chicago.
Roberto asked how much bandwidth would need to be dedicated to such
a network. Alain replied that 1 Mbps was recommended.
Roberto also asked whether the proposal required dedicated routers
instead of workstations. Alain replied that using workstations for
routing did not present a good image to NRNs and vendors. It was
possible the activity would not be taken seriously.
Juergen asked whether it was possible to interconnect with the 6REN
other than at the STARTAP. Alain replied it would soon be possible
to peer via ESnet on both the East and West coasts.
Mauro said the first step would be to set-up an IPv6 DNS Server.
Alain agreed and added the latest version of BIND could handle AAAA
records; both IPv4 and IPv6 could be used to make a query.
Nevertheless, this system would not scale in the long-term and a new
scheme had been proposed which dynamically builds DNS records.
Francis commented that DANTE probably needed to become a PTLA before
it could become a Sub-TLA. The draft guidelines for issuing IPv6
addresses required organisations to have involvement with IPv6
networks before they were granted registry status.
Christoph concluded this was an interesting activity. The goal of
the Task Force was to investigate technologies that may be used in a
production environment, and NRNs were already looking at migration
strategies from IPv4 to IPv6. He proposed the 6REN procedures should
be adopted for the TEN-155 IPv6 service. Simon was asked to mention
this initiative at the forthcoming RIPE meeting to determine the
level of support.
ACTION 2.4 - Simon Nybroe
4.6 IP over ATM
Roberto said he needed to test the performance of IP over ATM as
part of the management of TEN-155, but he asked whether anyone else
was interested.
Victor replied that SURFnet had conducted such tests over VBR
together with KPN and the University of Utrecht. A report would be
available shortly. He added that he would like to see similar tests
over lossy ATM.
Wilfried said he might be interested in this activity if VPs from
TEN-155 were being extended into NRN infrastructure.
Jean-Marc asked whether NRNs without Ascend switches could
participate. Christoph replied this may be possible.
Roberto was asked to draft a detailed proposal for this experiment.
ACTION 2.5 - Roberto Sabatino
4.7 ATM Signalling
Jan reported that KPN planned to start using ATM signalling from the
1st of June, and were currently defining the resources. A pilot SVC
service would be operated during September and October, with the
full service coming before the end of the year. The ATM signalling
experiment would investigate point-to-multipoint mapping, RSVP to
ATM mapping and PNNI.
Victor asked whether it was possible to investigate NNI. Jan replied
it would not be possible to define resources at that level. Rob
confirmed that KPN proposed to offer UNI, not NNI signalling to
TEN-155.
Jean-Marc asked whether a mesh of VPs could be established between
the core switches, with PNNI running over this. Rob replied this
might be possible, but their priority was to ensure production
services were not affected.
4.8 Policy Control
Victor reported that Leon Gommans from the University of Utrecht had
been asked to conduct a study of policy control. This would be
necessary to run things such as RSVP, differentiated services and
ATM switched services. A number of proprietary solutions were
currently available and these would be investigated, but some input
was required on what else was available.
Christoph asked Leon to send details of the policy control
mechanisms that were being investigated to the mailing list.
ACTION 2.6 - Leon Gommans
4.9 Route Monitoring
Simon Leinen was not present at the meeting, but he was looking for
comments on his proposal to investigate BGP analysis tools. This was
being done for SWITCH anyway, but it would be useful to know if
other NRNs were doing something similar.
Wilfried said Austria was currently conducting route monitoring, but
the tools being used were quite old. It would therefore be useful to
investigate the RATool set as a possible replacement. Roberto also
expressed DANTE's interest in this experiment.
Francis mentioned they had developed some PERL scripts to analyse
routing tables. These had been used for TEN-34 and would also be
used for TEN-155. Roberto was asked to make these available on the
WWW.
ACTION 2.7 - Roberto Sabatino
4.10 Flow-based Monitoring
Simon Leinen was not present at the meeting, but his proposal was
discussed.
Wilfried was concerned about interpretation of network statistics.
They could provide a general overview of a network, but they should
not be used to produce reports. Christoph however, said network
statistics were already being used in the UK and the Netherlands for
accounting purposes.
Dimitrios pointed out that NetFlow produced erroneous figures in
certain circumstances. This had currently halted their project to
evaluate traffic statistics.
It was agreed this experiment was useful to most NRNs, but it was
felt the aims and methodology could be better specified. Simon was
therefore asked to revise the proposal.
ACTION 2.8 - Simon Leinen
4.11 QoS Monitoring
Tiziana presented her proposal for monitoring QoS (http://www.cnaf.
infn.it/~ferrari/tfng/qosmon.html). This aimed to utilise network
monitoring tools to determine whether requested levels of service
were being met.
Jean-Marc said that MIBs should be included in the list of
monitoring tools for this activity. Tiziana agreed to incorporate
these.
Dimitrios mentioned the RIPE Test Traffic Project was undertaking
similar work. Kevin however, said this project involved commercial
networks and he believed much of the data was confidential.
Jean-Marc felt that traffic measurement issues justified a separate
meeting. Kevin replied that TERENA had already established a Task
Force for this purpose, but it had not met for some time.
Nevertheless, another meeting could be arranged if there was
sufficient interest.
4.12 MPLS
Jean-Marc presented his proposal for testing MPLS (http://www.
renater.fr/jmu/QTP/mpls-desc.html). This aimed to survey the
existing implementations, gain experience of the technology, test
advanced features, and determine its scaleability over a European
ATM backbone. Cisco had agreed to loan the necessary equipment which
meant a list of requirements had to be made.
ACTION 2.9 - MPLS experiment participants
Christoph noted the proposal mainly concentrated on Tag Switching.
Jean-Marc replied that he was starting with Tag Switching, but other
implementations of MPLS could be investigated in a later phase.
Christoph went on to ask whether the NDA signed for the TF-TEN
experiments was still in force. This could be a problem as results
from TF-TANT activities were being made public. Jean-Marc replied
the terms of the NDA still applied, but some compromise could
probably be reached.
Dimitrios also asked whether such NDAs would affect multi-vendor
interoperability tests. Jean-Marc replied this issue would have to
be investigated.
Juergen asked when the MPLS specification would be published as an
RFC. Jean-Marc was not aware of the current status of this, but he
would put references to the drafts on the WWW.
ACTION 2.10 - Jean-Marc Uze
Mauro asked whether MPLS was tied to IPv4. Bob replied it would be
reasonably easy to port MPLS to use IPv6, but he was not sure when
this would happen.
4.13 VPNs
Victor said SURFnet had started to investigate some products that
allowed VPNs to be established over public networks. This
investigation was also linked to policy control as VPN users often
needed to connect via external services (such as dial-up).
Wilfried suggested that end-to-end encryption should also be
investigated as part of this study.
4.14 WDM
Victor said he was still investigating the possibilities of testing
WDM. Unfortunately, PNOs were generally not willing to provide dark
fibre, so complicated negotiations were necessary.
Juergen commented they had some experience of WDM over the Gigabit
test network provided by Deutsche Telekom. They were currently only
using three wavelengths, but they planned to increase this number.
The test network had recently been extended to Berlin and this
required the installation of seven amplifiers. These appeared to be
working well.
Christoph asked Juergen to send some information about their WDM
experiences to the mailing list.
ACTION 2.11 - Juergen Rauschenbach
It was agreed this activity should be dropped from the list of
experiments until further progress was possible.
4.15 SDH Issues
Victor was investigating the issues related to STM-4. Any findings
would be sent to the mailing list for discussion.
Christoph said this was an important issue and could be discussed at
future meetings, but it probably did not need to be an experiment.
This was agreed.
5. VENDOR INTEGRATION
Christoph said he needed to determine what equipment needed to be
procured for the testing activities. He therefore asked experiment
leaders to specify equipment requirements in their proposals by
mid-February. In addition, he asked everyone to produce a list of
the equipment they already have.
ACTION 2.12 - Experiment Leaders
ACTION 2.13 - All
6. DATE OF NEXT MEETING
The next meeting will be held on the 29th and 30th of March 1999
at NTUA, Athens, Greece.
7. ANY OTHER BUSINESS
Bob announced that Cisco planned to host a peering workshop on the
23rd and 24th of March 1999 in Amsterdam. This would be preceded by
a workshop on multicasting on the 22nd of March and would be
attended by Cisco engineers working in this area. Both workshops
were aimed at NRNs and Internet Exchange personnel, but attendees
would require an invitation.
Christoph mentioned that he had drafted some designs for a TF-TANT
logo. He asked participants to review these and let him know which
design they liked most.
8. ACTIONS FROM LAST MEETING
1.1 Kevin Meynell to rename the TF-TEN mailing list.
- Done.
1.2 Kevin Meynell to create a link from the TERENA WWW pages to the
TF-TANT WWW pages.
- Done.
1.3 Christoph Graf to give the TF-TANT WWW pages a 'neutral' URL.
- Done.
1.4 Robert Stoy to produce a multicasting experiment proposal.
- Done.
1.5 Telebit to produce a proposal for for establishing an IPv6
network over TEN-155.
- Superseded.
1.6 Mauro Campanella to send the URL of the MONARCH project to the
mailing list.
- Done.
1.7 All Experiment Leaders to produce a list of requirements by the
1st of December.
- Superseded.
981004-1 Victor Reijs to send URL of GigaPort Activity Plan to the
mailing list.
- Done
OPEN ACTIONS
2.1 Simon Leinen to contact Tiziana Ferrari to determine what
activities could be included in the RSVP to ATM SVC Mapping
Experiment.
2.2 Robert Stoy and Jan Novak to specify the TEN-155
multicast facilities.
2.3 Tiziana Ferrari to integrate DYNACORE project into the
Differentiated Services proposal.
2.4 Simon Nybroe to mention IPv6 proposals at RIPE32 to determine
levels of support.
2.5 Roberto Sabatino to draft a detailed proposal for the IP over
ATM experiment.
2.6 Leon Gommans to send details of the policy control mechanisms
under investigation to the mailing list.
2.7 Roberto Sabatino to make routing analysis scripts available on
the WWW.
2.8 Simon Leinen to revise the flow-based monitoring proposal.
2.9 MPLS experiment participants to specify their equipment
requirements.
2.10 Jean-Marc Uze to put references to IETF MPLS drafts on the WWW.
2.11 Juergen Rauschenbach to send some information about WDM
experiences to the mailing list.
2.12 Experiment Leaders to specify equipment requirements in their
proposals by mid-February.
2.13 All to produce a list of the equipment they already have.
Contact: nep@dante.org.uk
DANTE | TERENA
[March 1999]