Skip to main content

You're viewing an archived page. It is no longer being updated.


RIPE Meeting:


Working Group:




Revision Number:


Wednesday, 9 May 2007
11:30 - 15:30 (UTC +0100)

Chair: Fearghas McKay
Minutes: Susannah Gray RIPE (NCC)
Jabber Scribe: Chris Buckridge (RIPE NCC)

A. Administrative Matters

  • Welcome
  • Finalise agenda
  • Approve minutes from RIPE 53

B. Select a Scribe
C. IEEE Higher Speed Study Group (HSSG) Progress Update | Greg Hankins
D. IXP Updates
E. Euro-IX Update
F. Video Peering White Paper - Feedback | Bill Norton
Z. A.O.B.

1. Video Working Group - Cara Mascini (AMS-IX)

There were no questions.

2. 100GE Discussion - Henk Steerman (AMS-IX) and Greg Hankins (Force10 Networks)

Greg explained that Henk had already given the scheduled talk on Monday and so this slot would be used for discussion instead.

He asked the room who saw the need for 100GE in their infrastructure in the near future.

Roughly a third of the room said that they had a need.

Greg explained that there had been a resurgence in interest in 40GE as a proposed standard. He asked the room who saw value in a 40GE solution that would be available in the same timeframe as 100GE. No one in the room showed interest.

Henk added that those proposing a 40GE solution have the idea that it will mainly be used by those who deploy a large amount of servers. He asked the room why they would you not be interested in a 40GE solution which would be ready around 2010 - 2012?

Lorenzo Collitti (Google) commented that he didn't think it was enough and was concerned it would delay the adoption of 100GE.

Keith Mitchell (ISC) said that when you order a magnitude up in ethernet capacity, you are looking at two to three-times the cost. He continued that 40GE might make sense to some people if it's going to cost a lot less than 100GE. What about the cost point of 40GE versus 10GE versus 100GE? He added that he thought this was an important question to address.

Remco van Mook (NDIX) commented that, at the rate of 40 GE, it would take quite a while to get any useful contact out of the server.

Mike Hughes (LINX) added that a discussion had been held on the mailing list about 40GE. He continued that you can do this already using link aggregation. He said that he thought 40GE is a waste of time and a distraction from the real job.

Henk responded that those people proposing 40GE say that link aggregation doesn't work, especially if it is used on single server and that the distribution of traffic over links and link aggregation doesn't work well because you don't have that many variables.

Mike said that this proposer is in the wrong IEEE working group and should direct his comments and efforts to the link aggregation group.

There was general agreement.

Greg commented that he was trying to find people with experience with large server deployments and link aggregations. He continued that, so far, no one has said they want a 40GE server solution and no one has said lag doesn't work for servers at 10GE. He asked if anyone had any input.

Bill Norton (Equinix) asked if 40GE would be on the market quicker than 100GE.

Greg responded that both are technically feasible. The feasibility has been proven from an optical layer and from an ASIC component layer by the study group. There's no real reason why 100GE can't be done.

Henk added that the time-scale for 40GE is similar to 100G and that the limitation is the standards process which takes 2-3 years. He continued that it might be interesting to have 40GE available next year but the whole process takes so long. 40GE can't be used as an intermediate step to 100GE.

Bharat Ranjan (Microsoft) commented that Microsoft has 10GE connected servers and 40GE on the uplinks and that he did not see a need for 40GE on the server. 100GE on the uplinks is a definite plus.

Greg mentioned an "MSA" - multi source agreement. He explained that there is a 40GE optical MSA which doesn't have many members but there will probably be some changes in this group when the IEEE figures out what to do with 40GE and 100GE. He continued that he thought that 40GE is out of the IEEE and that there will probably be interest in the 40GE MSA. He explained that it will be a wave division multiplex solution. It will be a simple 40GE optic that muxes four 10GE ports together. From an architecture perspective, it will be link aggregated within switch or router. He asked the room if they saw any value in this.

Mike commented that Extreme did something similar at 1GE and it was awful. He advised not to make the same mistakes as it did. He added that it would need to be competitive against trying to do it with existing 10GE lag and a standalone WDM kit or 40GE lag and a standalone WDM kit. It could mean that money is spent for very little gain.

Greg agreed with Mike and said that a 40GE solution would cost approximately EUR 20,000 in optics plus the WDM equipment. He asked the room if a 40GE optic at half this price would be attractive, taking trade-offs such as manageability, less equipment and fewer interfaces to manage into consideration?

There was no real interest in the room.

3. Good Practices in IXP Documentation - Bill Woodcock (Packet Clearing House)

Bill Norton (Equinix) asked if it was possible to do the same thing by writing code to screen scrape peering DB.

Bill Woodcock responded that this could, presumably, be done.

Bill Norton continued that his observation was that, ever since he's been looking at this data, exchange point participation lists and contact information are out of date almost immediately. Rather than having every exchange point in the world each try to update, why not have the peering DB as the standard place for all this information to reside

Bill Woodcock asked Bill Norton if he was suggesting that IXPs should no longer maintain websites.

Bill Norton continued that the contact information specifically need not reside in 100 different IXPs when it is the same for all of them.

Will Hargrave (LONAP) commented that the problem with the peering DB is that is is updated by the members and is therefore not a full list. He continued that it would be great if IXP operators could interface directly with the peering DB and that this has been requested in the past.
Bill Woodcock stated that he had pointed out that if there's a standard way for IXP operators to publish this information, peering DB could import that data from those Internet Exchanges.

Kurtis Lindqvist (Netnod) commented that Euro-IX members actually have a standard interface for publishing this data on the Euro-IX website in the Euro-IX DB.

Bill Woodcock announced that, later this month, PCH will publish a revision to the mass netflow analysis toolset that was written in 2000. The new one will be open source distributable and hosted on PCH's platform at exchange point so that if people don't feel like installing it and the back-end DB that supports the analysis, they can get daily or weekly report back from PCH. He continued that PCH will provide an agreement that it doesn't share or save any information.

He continued that part of the idea is that, once three or more ISPs are doing this in any country, PCH will be able to publish aggregate information for that country about how much traffic is going to and from other countries. Obviously, the more data the more accurate. A lot of government and NGO people will be interested to see how much traffic is going between different countries as a percentage; you can't extrapolate to absolutes.

He added that the reports from the netflow tool show how much traffic is terminating in every other AS or how much would go through any other AS that you were to interconnect with directly, a sum of the traffic for that AS and also all of the traffic for your downstream AS. You can figure out how much traffic you could deliver through a potential new peer. This has been done in the past on a batch basis. People want on-going reports rather than disks.

4. Route Operators at Exchanges - Joao Damas (ISC)
Joao did not use slides.

Joao stated that anycast nodes are being deployed. The current map is on the ISC website. He continued that one node is missing on this map: one was recently installed in Oslo. The aim of the deployment is to get as broad coverage as possible and from the map it can be seen that there are sill some places that need more coverage.

He continued that ISC is working on that. He added that Europe still has room for few more. He told the room if anyone had any interest in helping host an F installation or any other letter in their exchange or know anyone else who would like to host, they should contact him.

He informed the room that ISC has signed agreement with Newstar to use the Netshield installation to start embedding some of these installations in ISPs for the ISPs' customers to use. None of these notices are live yet. - Kurtis Lindqvist
Kurtis did not use slides.

Kurtis explained that he worked for Autonomica, which runs the I root server. He continued that if you go to website there is a list of all the root servers and where they are located. He asked if anyone was interested in peering with I routes to contact him. Autonomica has announced quite a few anycasted TLDs so you will receive routes not only for the I routes but also for a number of ccTLDs around the world.

J root - Mark Kosters (presented by Fearghas McKay)

There were no questions.

5. Directive of the Council on the Identification and Designation of European Critical Infrastructure and the Assessment of Need to Improve Their Protection - Malcolm Hutty (LINX)
Malcolm did not use slides.

Malcolm explained that the European Commission is creating a program for the protection of European critical infrastructure. This includes infrastructure from all sectors. The plan is that there should be a European framework for ensuring that the kind of infrastructure deemed critical at a European level is protected. This will result in the creation of a new regulatory framework.

Malcolm continued that the under the current plan, any infrastructure that is deemed critical to two or more EU member states or to one member state if the infrastructure is based in another member state is in scope.

He advised anyone who thought they may have a critical infrastructure to pay attention to what is being planned in this European directive.

He added that if your infrastructure is deemed critical, you will be potentially opening your organisation up to a lot of regulation.

He also added that it is unclear at this stage how onerous the regulation will be, and it may vary from country to country.

He continued that the directive is in draft stage at the moment and added if people thought it might affect them, they should already be paying attention to the to the scope and potential obligations you may incur

There were no questions.

IXP updates:

6. Euro-IX Update - Serge Radovcic

Serge asked the room how many people represented an IXP. Roughly three-quarters of the room responded that they were representing an IXP. He asked if any of these people were not a member of Euro-IX. All were members.

7. AMS-IX - Steven Bakker

There were no questions.

8. De-Cix Internet Exchange - Arnold Nipper

There were no questions.

9. Linx - Mike Hughes

There were no questions.

10. MIX Update - Simone Morandini

Maurice Dean (Google) asked, regarding the VoIP project, whether members had to be licensed carriers. Simone responded that yes, they need a license.

Maurice asked whether the license he referred to regarding content was an ISP license which is no change. Simone responded that it was.

Mauro Magrassi (MIX) added that, in Italy, you need a license to operate an ISP. The government does not care whether you are a content provider or an ISP offering services to end users. This license also works for VoIP.

11. NaMex - Daniele Arena

There were no questions.

12. NDIX - Remco van Mook

There were no questions.

13. Netnod, IEX - Kurtis Lindqvist

There were no questions.

14. Nix-cz - Josef Chomyn

There were no questions.

15. Top-ix - Simone Arena

There were no questions.

16. VIX Update - Christian Panigl

There were no questions.