Minutes of 3rd TF-TANT Meeting
|
|
|
TERENA/DANTE TASK FORCE FOR TESTING ADVANCED NETWORKING TECHNOLOGIES
Minutes of the 3rd TF-TANT meeting held on the 29th and 30th of March
1999 at the National Technical University of Athens, Greece.
Kevin Meynell - Issue 2
PRESENT
Name Organisation Country
---- ------------ -------
Pantelis Balaouras GRNET/U.Athens Greece
Michael Behringer Cisco United Kingdom
Marc van den Bergh KPN Research The Netherlands
Zlatica Cekro VUB/ULB Belgium
Howard Davies DANTE -
Larry Dunn Cisco United States
Tiziana Ferrari INFN Bologna Italy
Leon Gommans U.Utrecht/Cabletron The Netherlands
Christoph Graf (Chair) DANTE -
Ivano Guardini CSELT Italy
Jean-Marc Haye Torrent France
Joop Joosten CERN Switzerland
Dimitrios Kalogeras GRNET Greece
Tom Kosnar CESNET Czech Republic
Cees de Laat U.Utrecht The Netherlands
Simon Leinen SWITCH Switzerland
Ladislav Lhotka CESNET/USB Czech Republic
Edward Meewis Univ. Twente The Netherlands
Vassilis Merekoulias GRNET Greece
Kevin Meynell (Sec) TERENA -
Jan Novak DANTE -
Simon Nybroe Telebit Denmark
Damir Pobric IAT-CNR Pisa Italy
Themos Rapsomanikis NTUA Greece
Juergen Rauschenbach DFN Germany
Victor Reijs SURFnet The Netherlands
Esther Robles RedIRIS Spain
Roberto Sabatino DANTE -
Guenther Schmittner JKU/ACOnet Austria
Yannis Siahos Upatras-Net Greece
Robert Stoy RUS/DFN Germany
Jean-Marc Uze RENATER France
Marcel Wiget Nortel Networks France
Bert Wijnen IBM The Netherlands
Wilfried Woeber ACOnet Austria
Alexios Zarras GRNET Greece
Apologies were received from:
Mauro Campanella INFN Milano Italy
Brian Carpenter IBM UK United Kingdom
Olav Kvittem Uninett Norway
Bernard Tuy CNRS France
1. APPROVAL OF MINUTES
The minutes of the TF-TANT meeting held on the 25th and 26th of
January 1999 were approved.
2. STATUS OF QUANTUM & TEN-155
Roberto reported the lines between Germany and Austria, Germany and
Italy, and Germany and Poland had been installed since the last
meeting. The connections between France and Spain, Austria and
Slovenia, and from Portugal to either the UK or the Netherlands,
were still being negotiated. Other connections being planned
included a 34 Mbps line between Tel Aviv and London, an additional
155 Mbps line from either London or Amsterdam to the US, and a 10
Mbps connection between Ireland and the UK.
Some interim peering arrangements had been made to resolve the
interconnection problems with commercial networks during the
roll-out of TEN-155. These however, were only valid for six months
and a tender was being issued to obtain more permanent
interconnections.
The problems experienced with the Ascend ATM switches under high
traffic loadings had been resolved by installing additional
interface cards. The Cisco 12000 router in the US had also suffered
performance problems and had been replaced by two Cisco 7515s.
Obviously this was a concern because the Cisco 7507 routers
currently used on the TEN-155 network, would eventually not be
adequate. Alternative solutions from Juniper, Nortel and Ascend were
therefore being investigated.
The TEN-155 Managed Bandwidth Service (MBS) was successfully used
for a demonstration during the launch of the EU Fifth Framework
Programme in Essen, Germany. It would now move into a Beta-testing
phase with the TF-TANT, DYNACORE and EDISON projects.
Wilfried asked whether it was possible to obtain native ATM from the
MBS. Roberto replied this was possible, but NRNs would also need to
support native ATM over their local infrastructure.
Cees asked how the bandwidth for the TEN-155 demonstration in
Germany had been arranged. He understood that users had to contact
their NRNs rather than DANTE. Roberto replied this was the general
policy, but if a local NRN did not support the MBS, it might be
possible for users to obtain their own connection to a TEN-155 PoP.
Nevertheless, each NRN could determine their own policy on this.
Christoph mentioned the EU had requested that funding
notices/disclaimers be appended to everything published by TF-TANT.
He would circulate the relevant text on the list.
ACTION 3.1 - Christoph Graf
Kevin however, thought this was only reasonable for formal QTP
deliverables as TF-TANT was a volunteer effort with only the
Chairman and Secretary being partially funded by the EU. The group
agreed with this view.
3. TF-TANT OVERLAY NETWORK
Roberto said the previous meeting had discussed whether a permanent
overlay network should be established over TEN-155 for the TF-TANT
experiments. He presented plans for such an overlay using 2 Mbps
VPs, which would allow circuits to be established as required. It
had been designed to evenly share traffic between PoPs, although
some PoPs obviously had to support more connections than others. The
meeting was asked whether such an overlay should be set-up, or
whether the group would prefer connections to be set-up on a
per-experiment basis.
Simon L said the reasons for establishing a complex permanent
overlay network no longer applied. It had only been necessary to
establish such an overlay for TF-TEN because obtaining VPs from the
JAMES project was complicated. This said, the network was then
always available when required.
Wilfried was concerned that adding and removing connections on a
regular basis might affect the production network. Guenther agreed,
and added that ad-hoc connections made coordination difficult at
both the international and local level.
Victor asked whether single VCs could be switched between VPs to
create specific sub-overlays. Roberto replied that KPN did not offer
this as a standard service.
Lladislav thought a decision should not be made until the
requirements of each experiment were known. Christoph agreed, and
said he would circulate a questionnaire about requirements on the
mailing list. The initial experiments would therefore have to make
use of ad-hoc overlays.
ACTION 3.2 - Christoph Graf
Lladislav asked whether it would be possible to install test
switches in the TEN-155 PoPs. Roberto replied the purchase of these
would have to be approved by the QUANTUM Policy Committee.
4. REVIEW OF EXPERIMENT DESCRIPTIONS
4.1 Differentiated Services
Tiziana had presented her proposal for testing Differentiated
Services (http://www.cnaf.infn.it/~ferrari/tfng/diffserv.html) at
the last meeting, but she thought it should be extended to include
investigation of bandwidth brokers as the IETF was doing a lot of
work on these. The experiment was scheduled to start in April or
May, but this was dependent on when IBM could loan some equipment.
In addition, it would be better to conduct tests over 4-8 Mbps
tunnels as TCP back-off could cause inefficient bandwidth
utilisation over smaller tunnels.
Cees thought that 2 Mbps tunnels could be used if the number of
assigned priorities were reduced to two. Tiziana agreed this could
be done if it was not possible to obtain more than 2 Mbps.
Roberto pointed out the proposal assumed the existence of a generic
overlay network which might not exist. In addition, it involved
Spain which was not yet connected to TEN-155. It would therefore be
necessary to set-up tunnels on an ad-hoc basis. Tiziana replied she
would revise the proposed topology.
Christoph agreed that bandwidth brokers should be incorporated into
the experiment. These could be tested in a second phase when they
became available.
ACTION 3.3 - Tiziana Ferrari
4.2 RSVP to ATM Mapping
Tiziana reported this experiment was scheduled to start in July when
SVCs were supported by TEN-155. In addition, she was hoping to test
equipment loaned by Nortel Networks and Telebit.
Michael suggested the Cisco equipment loaned for the MPLS experiment
could also be used for this experiment.
4.3 QoS Monitoring
Tiziana expanded on her proposal that was presented at the last
meeting (http://www.cnaf.infn.it/~ferrari/tfng/qosmon.html). The
experiment was scheduled to start in April or May, but effort-wise,
this was dependent on when the MPLS experiment finished. It was
hoped that RIPE and/or Surveyor traffic measurement equipment could
be utilised to monitor one-way delays. At a later stage, testing
over the QBone could also be carried out.
4.4 IP over ATM
Roberto presented his proposal for testing the performance of IP
over ATM (http://www.dante.net/staff/roberto/docs/1999/qtp/
RS-99-04.html). The aims were to identify the best configuration for
combining best-effort IP and other types of traffic, and to verify
that SCR was guaranteed in the event of congestion.
Some tests had already been conducted at the KPN laboratories in
Hilversum. Three workstations sent varying amounts of traffic to an
Ascend CBX-500 configured to support different SCRs for each.
Unfortunately, the results showed the switch did not load balance
correctly when the lines were operating at full-rate. Further
investigations were therefore necessary.
Christoph said the experiment proposal did not describe the tests in
Hilversum. Roberto agreed to incorporate these.
ACTION 3.4 - Roberto Sabatino
4.5 Flow-based Monitoring
Simon L reported he had received a lot of feedback about this
project. A lot of NRNs appeared to be working in this area, or were
at least conducting investigations.
He was currently collecting pointers to relevant software and other
similar projects (http://www.switch.ch/tf-tant/floma/). The next
stage would be to install a few packages and start to analyse
traffic streams; mainly to obtain information about usage patterns.
Whilst a testing environment needed to be defined, this did not
directly require hardware and software resources from TEN-155.
4.6 RSVP
Simon L said there did not appear to be much interest in this
experiment and proposed to amalgamate it with the RSVP to ATM SVC
mapping experiment or scrap it completely. He did not consider it
worthwhile to conduct more tests under local area conditions.
Christoph asked whether anyone was planning to use RSVP, and there
was no response. Victor however, suggested the interaction between
DiffServ in a core network with RSVP at the edges should be tested.
He would send the URL of a document about this to the mailing list.
ACTION 3.5 - Victor Reijs
It was agreed any further work in this area should be conducted in
the context of the RSVP to ATM mapping experiment.
4.7 Multicasting (IP and ATM)
Robert presented his plan for testing native multicasting over
TEN-155. A number of independent multicast domains would be
established to allow PIM Sparse Mode, MBGP and BGMP to be tested.
Cisco equipment running IOS 12.x would therefore be required in
Austria, Germany, the Czech Republic, the Netherlands, Italy and
Spain. It was hoped this could be established before the IETF in
Oslo.
EURESCOM were also interested in collaborating with TF-TANT. This
consisted of twenty-one European PNOs who wished to investigate
multicasting amongst other things. They were not certain they could
obtain a suitable test infrastructure and were interested in using
TEN-155.
Christoph had some reservations about this type of collaboration. He
did not think it was acceptable for commercial companies to use
TEN-155 for their own research, especially if NDAs were involved.
Nevertheless, if they managed to obtain their own multicast network,
it should not be a problem to connect it to TEN-155. Wilfried added
that commercial companies should be welcomed, but on the
understanding that everything was in the public domain.
Wilfried (?) also expressed concerns about the impact of native
multicasting on the production network. Simon L however, said these
arguments had prevented the introduction of new technologies on
TEN-34. New technologies would have to be introduced eventually, and
TEN-155 presented an ideal opportunity to do this.
Christoph suggested approval for setting-up the multicast network
should now be sought from the QUANTUM Policy Committee. Roberto
agreed to make the necessary representations.
ACTION 3.6 - Roberto Sabatino
4.8 IP Version 6
Simon N reported that ACOnet, CESNET, INRIA/G6, DFN, GRNET, INFN,
SURFnet, Uninett, RedIRIS, SWITCH, DANTE and Telebit had expressed
interest in this activity. The intention was to build a
pre-production native IPv6 network that could be used for realistic
end-to-end experiments, gaining operational experience, and
identifying areas that required further development. Hopefully, it
would also generate interest in production IPv6 networks.
The plan was to run native IPv6 on SVCs over TEN-155. The
participating NRNs would provide transit to end-users and would
connect using either IPv4 or IPv6. This would allow tunnelling to be
tested. A connection to the 6Bone would also be provided.
Simon L said there was a lot of interest in this activity and he
proposed something should be established before the Oslo IETF in
July. This was agreed.
Christoph noted the use of SVCs was proposed, but they would not be
available until June at the earliest. Simon N replied they were not
a prerequisite and PVCs could be used if necessary.
Wilfried asked where the Telebit equipment would be located, and who
would be managing it. Christoph replied it would probably be located
in a KPN PoP in either Amsterdam or Frankfurt. DANTE would undertake
the management with assistance from Telebit.
Tiziana asked whether a connection to the 6REN was still planned.
Simon N replied this was not an immediate priority, and in any case,
TEN-155 was not yet connected to the STARTAP.
4.9 ATM Signalling
Jan reported on the introduction of ATM signalling on TEN-155. The
service had to be specified before the 30th of April, and
contributions to the process were currently being sought. KPN
planned to start using ATM from the 1st of June, and would be
conducting acceptance tests for approximately one month. Once these
were concluded, DANTE would conduct its own acceptance tests until
the end of September. If these were successful, SVCs would be made
available to MBS projects.
The introduction of SVCs should not affect the production network as
additional interfaces were being installed on the Ascend switches.
Both UNI 3.1 and 4.0 would be supported, using E.164 AESA
addressing.
In order to conduct the acceptance tests, DANTE were looking for
participants with Cisco LS-1010 or Fore ATM switches. Dimitrios,
Jean-Marc U, Simon L and Guenther/Wilfried volunteered for this.
Juergen commented that most NRNs used, or planned to use, DCC-based
rather than E.164-based addresses. Nevertheless, these could be
mapped relatively easily if necessary.
Guenther asked whether PNNI would be supported. Marc replied that
Ascend did not recommend its use, but it could be offered in
principle. It was therefore agreed that KPN would be asked whether
they would provide PNNI.
ACTION 3.7 - Guenther Schmittner
4.10 Policy Control
Leon reported this activity was currently in the research phase
which was scheduled until June. This aimed to identify the scope of
the work, acquire information on standards and products, and to
define testing procedures. The next phase would be to conduct
testing of different protocols (e.g. COPS, DIAMETER, PFDL), service
access models (host, user or application), user authentication (e.g.
Chipcards), and finally accounting mechanisms.
Marc asked whether any policy control products were currently
available. Leon replied he knew of a couple of access control
applications and a bandwidth broker implementation. He asked people
to forward details of any other known applications.
4.11 Route Monitoring
Simon L was currently investigating route monitoring software. The
next stage would be to install a few packages on a workstation on
TEN-155 to investigate their usefulness. Depending on the security
policy of TEN-155, the routing data could then be made available to
interested parties.
4.12 VPNs
Victor said RENATER was planning to implement VPNs, and this
activity should be postponed until they became available.
Nevertheless, some work could be conducted in the context of the
Policy Control experiment.
Dimitrios asked whether VLANs and Emulated LANs would be
investigated as part of this activity. Victor replied this was
planned, but they were not the highest priority initially.
4.13 SDH Issues
Victor said this activity had not progressed since the last meeting
due to lack of information. He would however, draft a document
expressing the concerns of the research community about STM-4c.
ACTION 3.8 - Victor Reijs
4.14 WDM
Victor said investigations were still ongoing. A number of WDM
networks were being planned, but the technical details were not yet
available.
Juergen gave a presentation on the two Gigabit testbeds in Germany.
One was located in the Nord-Rhein-Westfalia region with a capacity
of 2.5 Gbps. This used ForeRunner ASX-4000 switches, but did not use
WDM.
The other testbed ran between Munich and Erlangen, and had recently
been extended to Berlin. This was based on WDM and currently used
three wavelengths (as only three transponders were provided per
MUX), although it could potentially support eight. One channel was
used to connect three ASX-4000 switches, one channel was used to
connect three Ascend GX-550 switches, with the remaining channel
being used for further tests such as IP over WDM. The ATM switches
had STM-16 (OC-48) interfaces, with applications using STM-1/4
accesses.
Further developments would lead to the use of increasing numbers of
wavelengths (up to 48 at present) with higher data rates (10 Gbps).
New features like protection switching on the optical layer would
also be available soon.
4.15 MPLS
Jean-Marc U reported that ACOnet, CERN, CESNET, DANTE, DFN, GARR,
GRNET, KPN, RedIRIS, RENATER, SURFnet and the University of Namur
had expressed interest in participating in this experiment. Cisco
had agreed to loan some LS-1010 switches and 7200/7500 series
routers, whilst Netcom would loan some measuring equipment for this.
The experiment would be divided into four phases: Phase One was
scheduled for 15/4 to 10/5 and would consist of basic functionality
tests. Phase Two between 11/5 and 17/5 would investigate VPNs. Phase
Three between 18/5-24/5 would examine RRR. Finally, Phase Four
between 25/5 and 31/5 would test QoS, DiffServ, OSPF and BGP
routing, VC merging and intelligent discard, and performance of UDP
and TCP traffic.
Each experiment participant was asked to confirm their
participation, arrange ATM connectivity to TEN-155, complete the MBS
questionnaire, put a diagram of their local Tag architecture on the
Web, specify a delivery address for the loan equipment, and sign the
Non-Disclosure Agreement. Jean-Marc U would propose the backbone
infrastructure and VPN set-up, circulate the configuration document,
and organise a Smartbit tutorial in Paris. In the meantime, Cisco
were asked to check the hardware configurations, determine software
availability, and provide feedback on the test plan.
ACTION 3.9 - All MPLS experiment participants
ACTION 3.10 - Jean-Marc Uze
ACTION 3.11 - Cisco
CERN, CESNET, GARR, GRNET, KPN and SURFnet confirmed their
participation in the experiment. ACOnet and DFN did not currently
have local access to the MBS, whilst DANTE did not have the
manpower. DANTE were willing to install equipment in the TEN-155
PoPs, but someone else would have to manage it.
Marc asked whether it was possible to participate using a Cisco 4500
series router. Jean-Marc U was uncertain whether this supported
MPLS.
Guenther asked how the bandwidth for this experiment should be
specified. Jean-Marc U replied it should be in megabits per second,
although configuration would actually be done in cells per second.
Bert thought the number of participants in this experiment would
cause logistical difficulties. Christoph replied that many NRNs were
considering using MPLS, and they needed to gain experience.
5. VENDOR INTEGRATION
5.1 Torrent
Jean-Marc H gave a presentation on the Torrent IP9000 Gigabit
router. This was currently being beta-tested, with production
planned for the fourth quarter of 1999.
Two models were available: an 8-slot with a 10 Gbps backplane, and a
15-slot with a 20 Gbps backplane. These could accept a combination
of DS-3, OC-3/12, Fast Ethernet and Gigabit Ethernet interfaces. As
each physical port maintained a copy of the routing table, this
allowed fast routing decisions to be made.
The IPAction operating system was based on BSD Unix and currently
supported RIP1/2, OSPF2, ISIS, BGP4, IGMP, DVMRP, PIM, RSVP, NTP,
AAA and TACAC. DiffServ support was due in the summer, with MPLS
support later in the year.
Christoph asked whether Torrent wished to collaborate with TF-TANT.
Jean-Marc H thought this would be of interest to them from the
summer onwards.
Juergen asked about IPv6 support. Jean-Marc H replied this was not a
high priority at the moment.
5.2 Telebit
Simon N confirmed that Telebit would participate in the IPv6
activity. He had presented a proposal earlier in the meeting.
5.3 Nortel Networks
Marcel said that Nortel Networks were interested in the MPLS and
DiffServ experiments. They would be willing to loan equipment and
provide support if necessary.
5.4 IBM
Bert said IBM proposed to loan five 2212 and two 2216 routers for
the DiffServ experiment. They aimed to get some feedback on the
product line, so they would also be providing extensive support from
the UK. The loan still had to be approved by their research grant
department, but this was expected shortly.
IBM equipment did not currently support MPLS or RTFM, but it did
support IPv6. The AIX operating system also supported IPv6, so there
might be some scope to participate in the IPv6 testbed. In addition,
they were interested in the Policy Control activity.
5.5 Cisco
Michael said Cisco would be providing general support for TF-TANT,
with additional support for certain experiments. A dedicated contact
person had been appointed, engineering staff would be available, and
the group would have access to technical information and beta
software releases. The Cisco Laboratories in London and Paris could
also be used for testing, but they were quite heavily booked.
Finally, equipment was being loaned for the MPLS and DiffServ tests.
Larry then gave an appraisal of the TF-TANT Experiment Programme:
Differentiated Services - Very important, but timescales were
optimistic. Shaping at AS boundaries should also be investigated.
RSVP to ATM mapping - Not much demand for this. Requires Cisco IOS
12.0(3)T.
QoS Monitoring - Very important. Refer to work by IPPM, RTFM, CAIDA,
CSELT and UCLA.
IP over ATM - Medium importance. ABR was promising for best-effort
traffic.
Flow-based Monitoring - Important.
Multicasting (IP and ATM) - Important, but UCL should also be
involved. Please update obsolete Web pages!
IPv6 - Cisco might be interested in loaning equipment.
ATM Signalling - Relevant to TEN-155.
Policy Control - Very important, but Sweden should also be involved.
Route Monitoring - Did not appear to be much support, especially
from DANTE who originally proposed this activity.
VPNs - Could be merged with MPLS activity.
WDM - Cisco was very interested in this, but conducting tests will
be difficult and expensive.
SDH Issues - The problem will probably be solved in six months time.
MPLS - Extremely important for Cisco. Already loaning equipment.
6. DATE OF NEXT MEETING
The next meeting was provisionally arranged for the 17th and 18th of
June 1999 at INFN Bologna, Italy. This was dependent on the EU
Telematics for Research Concertation meeting being held on the 14th
of June, rather than the 18th as originally proposed.
7. ACTIONS FROM LAST MEETING
2.1 Simon Leinen to contact Tiziana Ferrari to determine what
activities could be included in the RSVP to ATM SVC Mapping
Experiment.
- Ongoing.
2.2 Robert Stoy and Jan Novak to specify the TEN-155
multicast facilities.
- Ongoing.
2.3 Tiziana Ferrari to integrate DYNACORE project into the
Differentiated Services proposal.
- Done.
2.4 Simon Nybroe to mention IPv6 proposals at RIPE32 to determine
levels of support.
- Done.
2.5 Roberto Sabatino to draft a detailed proposal for the IP over
ATM experiment.
- Done.
2.6 Leon Gommans to send details of the policy control mechanisms
under investigation to the mailing list.
- Done.
2.7 Roberto Sabatino to make routing analysis scripts available on
the WWW.
- Done.
2.8 Simon Leinen to revise the flow-based monitoring proposal.
- Done.
2.9 MPLS experiment participants to specify their equipment
requirements.
- Done.
2.10 Jean-Marc Uze to put references to IETF MPLS drafts on the WWW.
- Done.
2.11 Juergen Rauschenbach to send some information about WDM
experiences to the mailing list.
- Gave a presentation during the meeting instead. Done.
2.12 Experiment Leaders to specify equipment requirements in their
proposals by mid-February.
- Ongoing.
2.13 All to produce a list of the equipment they already have.
- Done.
OPEN ACTIONS
3.1 Christoph Graf to circulate EU funding notice/disclaimer on the
mailing list.
3.2 Christoph Graf to circulate a questionnaire about MBS
requirements on the mailing list.
3.3 Tiziana Ferrari to incorporate bandwidth brokers into the
Differentiated Services experiment proposal.
3.4 Roberto Sabatino to describe laboratory tests in the IP over
ATM experiment proposal.
3.5 Victor Reijs to send URL of document discussing the use
DiffServ in conjunction with RSVP to the mailing list.
3.6 Roberto Sabatino to obtain approval for the TEN-155 multicast
network from the QUANTUM Policy Committee.
3.7 Guenther Schmittner to ask KPN whether they will support PNNI.
3.8 Victor Reijs to draft document expressing the concerns of the
research community about STM-4c.
3.9 All MPLS experiment participants to arrange ATM connectivity to
TEN-155, complete MBS questionnaire, put diagram of local Tag
architecture on the Web, specify delivery address for loan
equipment, and sign Non-Disclosure Agreement.
3.10 Jean-Marc Uze to propose backbone infrastructure and VPN
set-up, circulate the configuration document, and organise
Smartbit tutorial.
3.11 Cisco to check the hardware configurations, determine software
availability, and provide feedback on test plan.
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.
2.12 Experiment Leaders to specify equipment requirements in their
proposals as soon as possible.
Contact: nep@dante.org.uk