Minutes of the 3rd TF-NGN meeting 2-3 April 2001 CESNET, Prague, Czech Republic

Valentino Cavalli, John Dyer, Issue 3
 

Table of Contents

•     Attendees
•     Apologies
• 1.  Welcome
• 2. TEN155/Geant Update
• 3. Premium IP service
• 4. Over-provisioned network performance analysis
• 5. AF-based Services, Traffic Shaping
• 6. Availability of European test bed
• 7. IPv6
• 8. GCS and MPLS
• 9. Measurement
• 10. Optical networking
• 11. Improved Multicast
• 12. Date of  Next Meeting
• 13. Any other business
• 14. Actions from last meeting
• 15. Open actions

Attendees

Apologies

Online presentations: http://www.terena.nl/task-forces/tf-ngn/presentations3.html
 

1. Welcome

The minutes of the 2nd TF-NGN meeting were approved without comments. The revised list of action is appended below. Overall it was decided to remove from the list all actions from previous meetings referring to ongoing activities in the current work items.
 
 

2. TEN155/Geant Update

The GEANT procurement was at the stage of short-listing connectivity providers/ network managers. Dante was currently negotiating prices with them prior to the selection of the providers. Howard mentioned an issue for the GEANT policy committee scheduled to meet on 11 April, with the presentation of two alternative scenarios: starting with 2.5 Gbps or starting immediately with 10 Gbps in the core. August 2001 is the date envisaged for the installation of the first circuits.
 
 

3. Premium IP service

A major issue in developing the IP Premium service is to understand which admission control parameters need to be measured, policed and hence then be included as part of the SLS or SLA. The ideas that he proposed as a starting point for discussion included:

• IP source address (an issue if you have a firewall, cache or other such device that hides the end system IP address)
• IP source and destination address
• TOS content
• Others

Other open issues regard the end-to-end propagation of the SLA and the support for asymmetric SLS. The issues involve a decision on a model for local vs global agreement between the bandwidth provider and the customers. To provide end-to-end service policing is needed at the NRN level and Mauro argued in favour of supporting end-to-end agreement. On the technical side shaping and aggregation need to be investigated because shaping maybe lost when packets are aggregated to higher link speeds. It was remarked that end-to-end measuring of ipdv would give insight on the need for shaping.

On Tuesday, before the end of the meeting the open issues were resumed and discussed again with the aim of reaching consensus about the solutions suggested by Mauro. There was some discussion with several people thinking that IP addresses would be too difficult and in any case unnecessary. The relationship between the SLS inside and outside the boundary is complex and there has not been sufficient work done to understand it. Many thought that TOS content would be a good starting point, but it was agreed that there should be a discussion on the email distribution list in advance of the next meeting. It was suggested to look carefully into the model of collaboration between the NRENs and GEANT, defining needs and service specifications. The suggestion of using the SEQUIN questionnaire as a basis on which to start thinking about a framework was accepted.

ACTION 3.1 Howard, Mauro to follow up on the definition of service specifications for the premium IP service.
 
 

4. Over-provisioned network performance analysis

It was generally agreed that it was very difficult to understand what "over-subscribed" actually means. Howard Davies asked what are the measures that would be used to decide when a net is either over or under provisioned. Simon Leinen said that when he wants to decide when a net was over provisioned, he would send a 1000 pings or so and look at the return times. If they all arrived back spaced as sent, then there are no queues etc, so it is fairly reliable measure that it is adequately provisioned.

The question being raised by Dimitrios is whether or not one can get a more discriminating measure by using something more complex than simple manual ping tests.
 
 
 

5. AF-based Services, Traffic Shaping

ACTION 3.2 Michal to stimulate investigation on usage of Linux or OpenBSD platforms for traffic shaping.

Further tests configuring the burst length and the inter burst time on the CISCO 7500 were undertaken with different packet sizes recording packet size, throughput and CPU load using three types of streams:

• perfectly shaped traffic at 5 Mbps
• bursty traffic with an average 5 Mbps containing 0.1 sec gaps between bursts
• traffic at up to 20 Mbps burst with much longer gaps giving an overall average of 5Mbps

Tiziana concluded by saying that this work is a starting point on understanding the performance of the shaper. What is needed now is to see the effects of aggregating 100-200 streams and the effect of shaping and queuing on bursty traffic.
 
 

6. Availability of European test bed

• What Test-beds currently exist or are about to exist
• What would we need to get access to them
• What bits are missing, equipment / open loops etc.,

GARR-G, Italy. GARR has a triangular test-bed of 2.5 Gbps links connected with three CISCO 12000 routers. The test-bed is planned to be operational during mid-April 2001. The test-bed and can be made available for SEQUIN and TF-NGN testing provided this can be scheduled into the internal GARR-G requirements. In addition to the CISCO 12000's GARR-G have some switches that can be used to tag and police traffic before entry into the CISCO routers.

SuperJANET4, UKERNA, UK. The SuperJANET4 network consists of eight core POPs each with a GSR. Four of the POPs in the centre of the network will have additional fibre links that will be dedicated to test bed activities and should be available from summer 2001. UKERNA will put three CISCO routers on the test bed which should be available for testing purposes by groups such as TF-NGN.

GEANT, DANTE. Howard said that the provision of any native test-bed was not at the top of their priority list but they will give it serious thought. Any decision about an international test-bed will have to be taken by the policy committee. Howard went on to say that DANTE has been in discussion with CISCO regarding the CISCO EU-Project proposal in the context of IPv6.

ATRIUM, France. Atrium is an IST project involving France Telecom, Alcatel, University of Liege and the University of Namur. It will consist of a triangle of 2.5 Gbps links equipped with Alcatel 7770 routers (which are high end core routers). DANTE has discussed how to interconnect to ATRIUM and has concluded on the basis of local-loop costs that connection through BELNET and RENATER would be the most cost effective solution. It is a reasonable working assumption that ATRIUM will be open to proposals for tests from TF-NGN.

POLAND. The polish academic network has installed an optical test bed based on Alcatel WDM technology. The total link path length is about 180 km and is open to TF-NGN testing proposals.

TERENA. John Dyer of TERENA said he had nothing to add to this list and had nothing else he could mention.

The tests to be undertaken, their duration and the networks involved were then summarized as follows.
 
 

Tests Activities	Duration	Networks
IP Premium Service	Temporary	GARR-G, PLAGE, Quaser(DE)-pt SEQUIN/NGN
MPLS	Temporary	PLAGE
IPv6	Permanent	Own Network Infrastructure
Shaping	Temporary	GARR-G or PLAGE
Monitoring	Permanent	Unknown
Lambda	Unspecified	Unknown

Victor expressed his opinion that the issue of lambda testing was rather complex and requires a great deal of understanding before lambda testing can be undertaken.

Valentino agreed to set up a web page listing the test beds, a contact point for each and give an indication of their purpose and objectives. Information about the test beds will be made available at http://www.terena.nl/task-forces/tf-ngn/testbeds.html

ACTION 3.3 Valentino to set up listing of TF-NGN test beds.
 
 

 7. IPv6

Probably the most advanced IPv6 work being carried out at the moment is being undertaken in the WIDE project in Japan.

The Internet2 IPv6 Working Group had recently met and been exploring the major drivers for IPv6 in the United States. Lack of address space and the need for 3rd Generation mobile phones are not seen as significant factors, to the extent that they are in Europe. The IPv6 test bed on Abilene spans 4 nodes using CISCO 7206 routers and tunneled connections - see http://www.internet2.edu/IPv6/. A major objective of the IPv6 test bed on Abilene is to sow the seeds for more widespread IPv6 uptake. Complementary activities include seminars and tutorial sessions, however a major issue for them in this respect is generating the workshop material. As a further act of evangelism IPv6 will be made available at every future Internet2 event. Current Internet2 IPv6 activities include the availability of IPv6 at the Giga-pops and the development of an engineering blue print of best practice for how campuses deploy IPv6.

Pascal explained that the Commission had organized an event on how to promote the deployment of IPv6 which had been attended by approximately 20 delegates. He expects a number of IPv6 project proposals for the current open call (closes 25th April 2001) to emerge as a result of this event. The slides from the event are available at: http://www.cordis.lu/ist/ka4/mobile/concertation.htm. Tim had attended it and presented his view about the main issues, not all of them raised at the meeting, that still need to be resolved:

• Address allocation methods needs /29
• Last mile technology ADSL, CATV, ISDN and dialup
• Number of machines that come out of the box that support IPv6 is low
• Middleware - as a means to support Grid access
• Some of the 3G operators are less willing to make a large investment (after buying licenses)

Joop Joosten of CERN mentioned that they have been using IPv6 addresses from QTP, TF-TANT and own prefixes. CERN has two routers connected by an OC3 line from CERN to Chicago, peering with TEN-155 IPv6 services, 6TAP and with Hunter Electric Internet Services (a company in California). A major issue identified by Joop is that whilst it is relatively easy to set up IPv6 connections between remote sites, distribution to end user locally is more of a problem with large numbers of legacy networking elements that have to be migrated. At CERN, they have significant numbers of Cabletron routers that do not support IPv6, they have many firewalls and security issues that have to be taken into account and indeed that name servers do not support IPv6 requirements. So far they have had little success in getting implementations.

Christian Schild of JOIN-University of Münster, Germany has been trying to get funds for five dedicated IPv6 routers to be co-located on the DFN backbone with the service IPV4 routers. His schedule is to get these available by November 2001. He announced that he had some success in implementing IPv6 between two between a Cisco4500 and a Telebit TBC2000 routers providing IPv6 tunnels to the universities. The two routers run native IPv6 between themselves over a 2m cable. The interesting part in this connection is that it is some kind of a little production network. The Telebit is serving all the 6bone connections and leaf sites (3ffe:400::), and the Cisco serves the network with DFNs production addresses (2001:638::), connecting all the DFN customers and leaf sides who already have a production prefix.

Stig Venaas of Uninett, Norway has had a native IPv6 network running for two years amongst four routers over a dedicated separate network that has recently been enhanced to STM-1 lines. His planned development includes the installation of a router in Stockholm that will run native IPv6 for peering between the Nordic networks and elsewhere (DFN, SWITCH and the Slovakian networks were mentioned). Stig was quite concerned regarding the need to build a new coherent tunnel infrastructure for Europe and the rest of the world. He reported that sometime he sees his traffic going to Japan and then back to the US, so something needs to be done.

Tim reported that UKERNA has an ongoing 6-month IPv6 study looking at address assignment, mobility issues etc. [see Tim's URL for details http://www.ipv6.ac.uk/gtpv6/] Chris Edwards from the University of Lancaster is a partner in the BERMUDA2 project from UKERNA which is studying IPv6 deployment issues along with UCL and Southampton University.  Chris is also interested in TF-NGN's QoS work as he is on the UKERNA Think Tank which is looking at
bringing Quality of Service onto the SuperJANET4 network in the UK. Tim mentioned that Lancaster has been busy with commercial partners such as Microsoft, CISCO and Orange in developing various IPv6 stacks including some work on IPv6 for mobile applications.

The Spanish RedIRIS network has a tunneled IPv6 network with a core router located in Madrid. They have generally been using CISCO equipment and Solaris workstations. During March 2001, a tunneled IPv6 connection with static routes to Mexico was introduced. RedIRIS are currently in discussions with RIPE to obtain allocations of IPv6 address space and also in discussions with a Spanish SME to use them as the national registry.

Simon Leinen said that SWITCH was in a very similar situation as RedIRIS. The SWITCH head office is IPv6 ready and they are thinking about putting IPv6 into the national registry. Simon has tried to obtain a new router (a recycled 7500) to support the IPv6 migration. Alex is now working for the NOC has done a lot of DNS IPv6 development and will be working more generally with IPv6 in the future.

Tim Chown said Joao da Silva of the European Commission has undertaken a survey of the EU IPv6 projects and produced a 30 page document on his findings. A major ambition must be to provide interconnectivity between all these projects, rather than leave them isolated. In closing the section of his report, he mentioned a 45Mbps link that is available to connect from Japan to UCL and from there into the European academic network. That could provide some temporary solution for IPv6 interconnection from Europe to Japan until November 2001.

Tim gave a brief update on the IPv6 work item list that now includes:

• Transition methods including IPv4 over IPv6
• DNS issues
• IPSec - Evaluation FreeSWAN/Pluto, KAME/racon - Tim invited anyone who wanted to test against these implementations to set up their own IPSec gateway and have a go.
• Study of issues in using IPSec - need to look at the requirements and usage of IPSec and clients
• Tim had some discussion at Southampton about what applications are needed for IPv6 - mainly peer to peer. (The Internet2 hold this view also) Unified messaging ICQ, AIM, video mail and ticker messaging,
• JAVA enabled architecture-independent networking
• Firewall Testing (ip6fw - FreeBSD stability disappointing), netfilter Linux working on improving IPv6, ipfilter

He brought to the Task Force's attention:

• RIPE document RIPE-196 which is a review of how IPv6 address allocation should be handled
• multi6 - new IETF WG to study multi-homing and to define best practice. There are no real standards at the moment.

IPv6 Test Bed Project Proposal
Graça Carvalho, CISCO

Graça gave an overview of the schedules for release of IPv6 in their major product lines from 800-12000 series platforms. These are summarized here, for more detailed information see the slides of her presentation, which are available at http://www.ipv6.ac.uk/gtpv6/
 
 

Phase	Capabilities	Release Details	Release Dates
Phase 1Early Adopters	Basic IPv6	Series 800-7500 platformsIOS 12.2(1) T	May 2001
Phase 2a	Backbone Deployment	CISCO IOS  12.2(3)T or 12.(4)T	Beta with selected group April 2001
Phase 2b	Backbone Deployment	12000 specific release12.0ST	After May 2001
Phase 3Enhanced Protocols	OSPFv3E-IGRPMobile IPv6IPSecIPv6 multicast	IOS 12.2	Estimated date CY 2002
P3 Enhanced services	IPv6 QoS, Stats and Tunnels (IPV4 over IPv6)Hardware acceleration o GSR Cat.6kEncapsulation DPT and Cable over DSLNetMan - SNMP over IPv6 MIB update to RFC 2851)IETF IPv6 enhancements will be incorporated in Phase3

CISCO are planning a pan-European IPv6 experiment with the objective of testing production CISCO IPv6 equipment (rather than prototype) in a large wide area test environment. CISCO is willing to put in a large investment of equipment and engineering resource into a test-bed venture. CISCO made it quite clear that they wish to be the prime contractor in any resultant EU-funded project which results out the proposal which will be submitted into the RN.2 part of the IST programme.

The nature of the work to be proposed is to build a "real production like network" without the use of overlay which in reality means that a dedicated physical infrastructure will be needed and this cannot be accommodated on the GÉANT network. Graça Carvalho has been talking to DANTE regarding the possibility of them providing some parallel infrastructure. Discussions on the issue are ongoing, however it is difficult to plan as there is no indication of how much capacity, between which points may (if any) be available. CISCO is looking for an infrastructure with 2.5 Gbps in the core around the central part of the network with perhaps lower bandwidths (STM-1?) at the periphery. The project would be split into four phases as follows:

Phase 1 (duration 3-6 months from start of project)

Introduction of an IPv6 Infrastructure over IPv4 tunnels (nothing new here, this is what is currently widely available), so the facilities should be available by the end of 2001.

Phase 2 (duration 3-6 months)

Build a dedicated link layer for native IPv6, so that it would not impact on any IPv4 traffic.

Phase 3 (duration 24 months)

Put in place a dual stack network that would require some inter-working capability or develop an IPv6 only network, which requires the migration of the stacks of all machines on the network. It is CISCO's very ambitious intention that some of this phase development could be beta tested in parallel with phases 1 & 2 before the end of the year 2001. Pascal Drabik from the European Commission applauded this timescale.

Phase 4

Definition of responsibility for making use of the results of the outputs after the formal EU-project has ended.

CISCO would also like to look at the issues involved in supporting IPv6 mobility and IPv4/IPv6 gateways. In this connection, it was noted that SONY has developed an IPv6 stack which is embedded in its Playstation2. SONY have also developed an IPv4/IPv6 gateway, and have consequently been identified by CISCO as a potential commercial partner in the project. CISCO pointed out that whilst Internet gaming might be considered somewhat trivial in the sense of research networking, the application is an extremely good example of something that can adequately stress the network for time-sensitive communications. The games community understand the issues of time-sensitivity very well.

The open challenges to be addressed by the project include the following list, however it was realized by CISCO that there would probably not be sufficient time or effort to address each and every item:

• Determining the potential future IPv6 router table size and growth rate.
• The IPv6 performance level needed for "real IPv6 users" using production equipment on production quality networks.
• Mobile IPv6 clients applications and supporting network requirement design.
• The requirements and comparative merits of using IPSec security versus the use of traditional style firewalls on IPv6
• Design of the IPv6 QoS requirements for 3rd Generation mobile networks
• IPv6 Network Management tools
• IPv4/IPv6 Transition tools
• Field Training
• IPv4/IPv6 and vendor Interoperability
• Other items not listed above

• Building a pan-European IPv6 test-bed
• Studying the issues involved in the migration from IPv4 to IPv6
• Monitoring the deployment and impact of introducing IPv6
• Evaluating the manageability of a large-scale IPv6 network
• The introduction of new services that make use of IPv6 facilities (for example security)

• A set of guidelines and cookbooks describing best practice in migrating from IPv4 to IPv6
• The development of IPv6 Internet exchanges and the issues of multi-homing
• Support for specific user communities (Wireless GRIDs and Time Sensitive application users were mentioned.
• The development of IPv6 network operations cookbooks
• The deployment of a large-scale IPv6 network by 2002

• CISCO themselves as prime-contractor
• Either SONY or IBM
• DANTE
• Members of the National Research Network Community. It was stressed by CISCO that there is a need for showcasing the output from the project.

Jürgen Rauschenbach of DFN-Verein, Germany thought from the technical perspective this was a good initiative, but was concerned that it focused almost exclusively on the development of the wide area facility or backbone. He said that from the DFN viewpoint, there should be increased input on the user communities. Howard suggested that this could only be achieved through NRN involvement which could bring universities into the project. Howard also pointed out that NRN's should be aware of the resource issues participation might bring such as provision of dedicated high-speed local connections between the institutions and the project backbone. Local-loop connectivity at high speed is often very expensive and potential partners should keep this in mind. CISCO said that whilst they are willing to put in the necessary equipment and provide adequate engineering support, connectivity would have to be funded through the project.

There was some concern that an homogenous CISCO based backbone would not reveal the interoperability issues associated with using equipment from different vendors. This is to be addressed by peering the project backbone with other experimental networks using other vendors equipment. Graça suggested that since the experimental backbone must be of production quality, it would be necessary to operate it in a production quality way and that a dedicated NOC would be needed. CISCO seemed keen to provide this function as part of their contribution.

Questions regarding the potential users and interconnectivity of the backbone were raised. Graça said that the network would be open for all other EU-funded projects to use as a test-bed and expected connectivity to be made to the US and to Japan. There was some discussion about the willingness of NRN's to commit their own resources to this project, however it was generally agreed that they will have to make that investment to move to IPv6 sooner or later, and since CISCO and the EU might make significant contributions to this initiative, the project represents a good opportunity. Pascal from the EU reinforced this by mentioning the agreement of European ministers at the Stockholm summit, pointing out that the message has gone back to member-states that they should adopt IPv6 at the national level as a matter of policy.
 
 

8. GCS and MPLS

In the first tests, a subpool at 100mbps and a global pool of 2.4 Gbps was used and heavy SP losses were observed. Herve said they were not sure if there is a problem with the DiffServ configuration as they also need to investigate queue usage. Graça gave the name of a CISCO contact in France who will be able to help with resolving this issue. Herve stressed the importance of matching the loan with sufficient support, a serious remark that was noted by CISCO. He went on to explain that there will be interconnection between the France telecom test beds (which is supported on Juniper equipment) and Plage to test Inter-domain MPLS issues. Herve also emphasized the reason for testing MPLS is not just to experiment with it, but also to do useful things that can in future support real services. Herve asked for feedback from the TF on issues that might need addressing for service support. Howard Davies said he would like to get the DANTE technical staff involved with this work to evaluate how to use MPLS effectively on GEANT.
 
 

9. Measurement

NETFLOW & TRAFFIC MEASUREMENT
Simon Leinen, SWITCH, Switzerland

Simon reminded the Task Force that the Passive and Active Management (PAM) 2001 conference would take place between April 23-24 in Amsterdam. The event includes presentations on deploying passive measurement in large networks. Whilst the programme is on the web site, full papers are not yet available for public access. Simon said that he found the Extreme Happy Network Tool (EHNT) also MHTG interesting, but was concerned that it may not scale to large nets and would most probably be suitable for campus networks.

It was reported that Standard Floe Export was discussed at the last IETF and as a result there will be a Birds-of-a-feather session at the London IETF. Discussion is currently being undertaken on the IETF RTFM email distribution. To subscribe to this list send a message to majordomo@auckland.ac.nz containing the single line : subscribe rtfm.

David Harmelin of DANTE reported on his software package that takes a few samples every few seconds to detect Denial of Service Attacks. It is available free from DANTE, but sites need to sign a license in advance.

Simon is working on SNMP-based routing and as a result needs to get access to as many NRN/TEN-155 routers as possible from Swiss test workstation (ws2.ch.ten-155.net). He would be grateful if TF members could ask their network operations for access to routers from his workstation so as to be able to grab data with the goal of experimenting with multi-domain monitoring and use of diagnostic tools particularly for multicast issues. If TF members would like to try the tools, ask Simon for access. As the multicast function is usually well instrumented with SNMP MIBS, it is possible to collect excellent information on what is happening. A web based interface to Simon's tools was explained in which it was added that the use of other aids like beacons can produce a comprehensive and automated picture. With multicast this is particularly useful as often problems are not bi-directional and as a consequence, intermediate routers must be checked individually.
 

QOS MONITORING and SLS auditing
Victor Reijs, SURFnet, The Netherlands & HEAnet, Ireland

Victor started his presentation by mentioning the E2E performance initiative from Internet2 which has the objective of trouble-shooting end-to-end performance issues for users experiencing difficulty with high performance applications running over the network. He thought that NGN should have some collaborative links with this group. He also mentioned the Internet2 QoS working group in this context.

His opinion is that we really have to understand the network requirements of applications and he would like to evaluate the response time and error rates that can be observed when running MPEG-4 streams over a Wide Area Network. Microsoft have an "almost MPEG-4 compliant" application and Dimitrios reported a company he knows with an application that produces a fully MPEG-4 compliant data stream that could be used to stress the network. The obvious metrics to measure are one-way delay and the availability of capacity on the links being used. Tools such as pathchar can be used to estimate the bandwidth on each hop but one-way delay is a much more difficult thing to measure. One needs to know accurately the time at each end of the link, the only available tool at the moment being the use of GPS, even then it is often difficult to obtain the satellite signals. The latest version of the RIPE measurement boxes uses CAT5 cabling from the workstation to antenna, which has improved results, but can still be an expensive installation.

It was suggested in discussions that maybe all POPs should be equipped with GPS and a time stamping capability. Victor thought that to have this facility at each University would be useful. Tijany reported that they are redistributing the GPS time to a limited number of PC workstation using NTP. They plan to compare times between the GPS and distant sites and will report the results back to TF-NGN.

ACTION 3.4 Tijany to report on the usage of the new generation RIPE measurement boxes.

Victor is in discussion with RIPE to get permission to introduce CHARIOT onto the RIPE boxes. RIPE are open to the idea, but there is some implementation problem in that it won't compile. For 200 end points, cost of CHARIOT is something like 20k GBP which is expensive.
 

10. Optical Networking

There are two major optical networking activities to be undertaken in TF-NGN. The first, Information Exchange has been successfully underway for some time and the reports at this meeting are a good example. The second activity of getting hands on experience, in particular experience of optical connectivity between countries has not yet started. Victor then handed over to speakers from the Task Force to report on their experiences to-date.

The Polish Optical Test bed has been in place since 2Q2000. It consists of about 180km dark fibre, which has been obtained from the Polish railway company and is used to link MANs in Poznan. The main goal of their activity is to evaluate both the technology and equipment from Packet Engines, Marconi and 3Com. The equipment installed can deal with 16 wavelengths but currently only five are available and three are in use. The three wavelengths are used for ATM at 2.5 Gbps, GigaEthernet and IP over SONET

In addition to the Poznan MAN fibre, there is a trunk link between Poznan and Wroclaw.

The ambitious Polish Pioneer programme was presented. The first pioneer fibres are due for installation in 4Q2001 as part of a project that will continue until 2004. The intention is to link all Polish institutions by fibre during this period. There will be a launch Pioneer conference in Poland on 24th April 2001 at which equipment from manufacturers including Alcatel, Microsense and CISCO will be demonstrated. It is hoped this will remain in place for sometime. Access to the Polish optical network by TF-NGN for testing purpose is possible.

In general discussion it was agreed that it would be useful to experiment with different lambda for different services (compared to one lambda for many services) to evaluate advantages and disadvantages of each approach. Other experiments would include the development of an optical VPN on demand pilot. Victor said that in the context of ATM, network operators had tried to give dedicated pipes to specific users, but IP addressing issues made it too complex. His view is that there is little difference in the solution between ATM and optical in respect of the addressing issues and the same problem will still apply. Howard was of the opinion that ATM offered the advantage of international standardization over optical technology in that not all optical equipment was capable of interworking, a particularly difficult issue across international and multiple management domain boundaries.

There was some discussion on the level and origin of error rates on optical networks. The Polish network has been in operation for more than one year, and although no formal BER measurements have been made, it does not exhibit any obvious error problems. It was agreed that they would undertake some formal measurements and report back to TF-NGN. The view of Graça from CISCO is that their equipment is unlikely to be the source of errors; it is more likely to be from the fibers. Dimitrios said that problems that they had seen in Greece had usually been seen as a result of high-temperatures and dust on optical connections. It was agreed that error rates should be no worse than 1 in 10-12. Graça said that CISCO has some information on error rates on optical networks than she would make available to TF-NGN. Dimitrios added that the initial design of optical networks is complex and should include consideration of influencing factors of not only power budgets, but also things like unexpected fibre characteristics, so designs should be made on the basis of the maximum number of lambdas that are to be used.
 

Active Optical Components
Dimitrios Kalogeras, GRnet, Greece

GRNET has been undertaking a series of experiments to gain familiarity with optical technology in advance of the migration of their network to optical technology. The initial tests had been taken out within the university, as the Greek Telco had been unable to give them access to the necessary fibers on their network. Equipment Fore, Marconi and Siemens have been evaluated, some of which relies on transponders, which are very flexible but also extremely expensive. The Marconi PMA-8 electro-optical device for instance allows for the adjustment of the transmission and receiver frequencies as well as giving an analysis of the BER. Dimitrios said his experience was that 10-11 was easily achievable and 10-12 was possible. He repeated his observation that the fibre link itself was unlikely to be the source of errors, but dust and temperature variations at the interfaces could be a big problem. He said in one case it had taken a month to stabilize the network after installation. Some of the more expensive devices allow reconfiguration of the services (for instance one can utilize different lambdas for different services) without physical hardware intervention. The changes can be achieved by soft reconfiguration via a workstation interface.

Siemens have supplied an ADVA device (ADVA is an OEM for Siemens), which is physically supporting around 8 lambdas with transponders interfaces in three classes: 10BaseFB, 100BaseFB, STM-n (where n can equal 1, 4 or 16). Since ADVA also supply OEM equipment to CISCO, interoperability between Siemens and CISCO is possible without too much trouble.

The main issues to be considered in choosing fibers includes the dispersion shift and level of attenuation, which governs the distance needed between regeneration stations. At a level of 3G regeneration of the SDH frame structure is needed at anything between 8-180 Km at speeds of 2.5 Gbps. If you are planning to migrate from 2.5 Gbps to 10 Gbps, then 10G should be used from the outset. Dimitrios pointed out that the DWDM standards do not specify one common frequency spacing between the wavelengths. 40, 50, 100 and 200 are all possible, so choosing equipment throughout the network that can support the same spacing is essential. Circuit protection and restoration is also another area of multiple standards with some manufacturers adopting SDH based techniques, other using proprietary methods.

Dimitrios explained the Request for Proposals on WDM Networking Equipment that they have issued which calls for a frame and bit rate independent fully optical service of initially 2.5 Gbps upgradeable to 10 Gbps. He has asked for the institution equipment to be able to support 32 lambdas. In the Athens MAN, which is an 80km pair of rings, the installation must be able to operate without the need for regeneration.

In closing the session on optical networking, Victor requested that a slot for optical networking be provided at the next TF-NGN meeting in order that RedIRIS be able to present the call for tender for optical networking that they have produced.
 
 

11. Improved Multicast

There was some discussion on how to deploy the tool more widely, however it was generally agreed that a large matrix of beacons might devalue its usefulness as the update times would increase and a large matrix display might be difficult to read. Members of TF-NGN were encouraged to use the tool and provide feedback.
 

12. Date of Next Meeting

The following meeting was tentatively agreed as last week of September in GREECE, however Graça says this clashes with the CSICO networkers conference and the audience agreed that it should be moved.

[Secretariat note: There is a large cluster of meetings around the end of September, beginning of October 2001, including Internet2, RIPE - consideration should be given to the period 10-19th October.]
 

13. Any other business

14. Actions from last meeting

1.3 Tiziana and Mauro to provide support to GRnet in the definition of tasks and test plan for the OvpN activity.
 - DONE

ACTION 1.6 Octavio, Tiziana, Tryfon and Mauro to coordinate testing with deadlines for the premium IP deliverable.
 - DONE

1.13 Robert, Tiziana and Lada to refine the workplan for QoS & Multicast, which has relations with both premium IP and multicast services but was decided to be kept separate from them.
 - ongoing

1.14 Dimitrios to work out a more specific proposal about IP VPN service and see what can be related to the premium IP service.
 - DROPPED, This plan will be covered in the GCS work item

1.15 Tim to fix the priority areas and responsibility.
 - DONE

1.16 Tim and Bernard to work out a draft plan for set-up of the IPv6 testbed before the next TF-NGN meeting; RedIRIS will participate by installing an IPv6 router with ATM card.
 - ongoing

1.17 bernard to send to mailing list a format being used to distribute IPv6 addresses to customers.
 - ongoing

1.18 Tiziana. Mauro, Roberto, Herve to finalise discussion about loan of  equipment.
 - ongoing

2.1 the MPLS group to check whether MPLS traffic should be tested in  isolation or mixed with BE.
- ongoing

2.2 Octavio to make plans for using PlaGE for AF-based testing.
- ongoing

2.3 Octavio, Tryfon, what equipment is needed to interconnect the  PlaGE and the GARR-G testbeds and to take into account impact of ATM connection for testing the queueing techniques which may not work on ATM
interfaces on high-end routers.
- ongoing

2.4 Tim, to gather information on the UK testbed and post it to the  list.
- partially done- no commitment from UKerna yet

2.5 Rüdiger to investigate what traffic testing kit is available from Abilene in 60 Hudson and how they can be used with SWITCH; Simon to verify the options from SWITCH's point of view in Geneva or Zurich.
- DONE

2.6 Simon, Tryfon and Mauro to make plans for international tests.
- DROPPED, plans not feasible in the short time available

2.7 Bernard to send revision of the RIPE 196 document.
- DONE

2.8, Roberto to distribute Beacon SW and documentation
 - DONE
 

15. Open actions

1.13 Robert, Tiziana and Lada to refine the workplan for QoS & Multicast, which has relations with both premium IP and multicast services but was decided to be kept separate from them.
 - Postponed to the agreement on the IP premium deliverable

1.17 bernard to send to mailing list a format being used to distribute IPv6 addresses to customers.
 - ongoing

1.18 Tiziana. Mauro, Roberto, Herve to finalise discussion about loan of  equipment.
 - ongoing

2.1 the MPLS group to check whether MPLS traffic should be tested in  isolation or mixed with BE.
- ongoing

3.1 Howard, Mauro to follow up on the definition of service specification for the premium IP service.

3.2 Michal to stimulate investigation on usage of Linux or OpenBSD platforms for traffic shaping.

3.3 Valentino to set up listing of TF-NGN test beds.

3.4 Tijany to report on the usage of the new generation RIPE measurement boxes.
 

| Home | Information | Conferences | Innovation | Technical | Library | News |