This news blog provides news about the e-IRG and related e-Infrastructure topics.


Global Network Architecture is strong believer in collaboration sharing capacity, costs for transatlantic connectivity

At the occasion of the recent news that research and education (R&E) networks in Europe and North America have joined forces to deploy links providing a total bandwidth of 740 gigabits per second (Gbit/s), we had the opportunity to interview René Buch, CEO of NORDUnet, one of the partners in this cross-atlantic collaboration that has been set up under the umbrella of the Global Network Architecture (GNA) initiative.

The recent Global Network Architecture (GNA) announcement is about collaboration, René Buch explained. Many national NRENs have their own capacity connection capacity but this used to be all non-coordinated. As soon as you were traversing the Atlantic or the Pacific, you had to do it on your own. If you wanted redundancy, you had to provide this on your own. All the NRENs were purchasing capacity on their own. Then, six NRENs decided to work together in the Advanced North Atlantic (ANA) Collaboration, which is the predecessor of GNA. They wanted to buy capacity in bulk from commercial providers to get a better price. The more you buy, the better prices you get: that is usually how it functions.

The six NRENs were able to manage a far better price for the transatlantic connection. Instead of having six individual systems that were all non-redundant, they asked themselves: why not have a whole group of networks connected as one system, so that if something goes down, it switches to the other parts and the partners can share the bandwidth and the costs. Instead of each going on their own, they all winded up in a solution where the transatlantic connectivity is much cheaper and much more resilient than the partners would have achieved on their own.

When you start collaborating with others, they might have other collaboration procedures and other sets of technical specifications, aligning those, that is what GNA is about. Instead of having to work all the time with bilateral agreements to share connectivity in a remote part of the planet, the NRENs have a system where they have agreed on the operational paradigm, on how to deal with provision, and on the sharing of costs. One of the things that the NRENs are also trying to do is when they have shared something, they do not split it between four or five NRENs, which is not very efficient. If you split 200 Gigabit/s in five parts, then each party has 40 Gigabit/s but one 40-Gigabit/s part is not sufficient. And remember that any bit that you have used is wasted for ever. You cannot save bits.

Instead GNA decided to provide a guaranteed bandwidth. With the additional bandwidth that GNA is inducing, it is possible to burst through this ceiling. There is always the guarantee to have a better service. This gives GNA the ability to offer capacity to the people who want to take it on a best-effort basis at a very affordable price and with new products coming in.

GNA is about collaboration and about operation standards. As an example, NORDUnet is the contracting partner of one of the transatlantic circuits but the operation is actually done by SURFnet. Because of the time differences for Canarie which has an exchange point in Montreal, NORDUnet has no 24/7 operations but it is possible to help them at a cost that is lower for everybody. This is about using resources and sharing capacity most efficiently. The aim is to get a broad consensus of how the NRENs work together so GNA can expand this from just being a few NRENs to a global collaboration with a global service-committed bandwidth. People have to declare what they want to contribute and ensure that they will make it GNA-compliant. It is a question of making transparent what GNA can use.

GNA wants to keep the entry level low but, on the other hand, it wants to be firm on what the model and the operational paradigm is. It will not be easy to create a global platform based on local needs. It does not make sense for NORDUnet to buy capacity to size America but other NRENs can bring capacity in NORDUnet's network. The cost of your footprint is actually smaller than when one tries to broker out. This is about defining in the community of how GNA wants to work together. This is what GNA has been doing on the technical and policy part: defining the parameters for best behaviour. GNA has now 300 Gigabit/s per second for the transatlantic part and for the committed users 200 Gigabit/s. So, there is 100 Gigabit/s that is not service-guaranteed.

How can GNA deal with countries who want to use this for free and not contribute? It is not that GNA per se wants money from them. If countries from the Middle East or Asia come to GNA and ask for 10 Gigabit/s committed on the transatlantic part, GNA counters with the proposition to have capacity from Amsterdam to somewhere in Asia for instance. One NREN invests in the transatlantic part which is cheap for the other one, and the other one invests in connectivity between Europe and India or Pakistan which is beneficial for the first one. This will keep the investments locally but gluing them on a global scale and adding them together on known parameters, agreed procedures and agreed cost sharing. This is what GNA is all about.

We live in a world where there will never be one partner who builds it all. Even if four or five NRENs will build a global array, you will only create resentment from the other NRENs that will feel like being left out. GNA wants to include as many people as possible in a global scenario. GNA now really has won international traction. Emerging local community NRENs are using GNA as a lever to be part of this development and get investments from their government for a research network that will help the entire community.

GNA is about sharing capacity, sharing costs and getting more out of it than when you would do it on your own.

René Buch said that it is pooling what GNA has, enhancing collaboration, enhancing purchasing power. By standing together, the NRENs become stronger and get more capacity and provide better services for their community for the same amount of money. Research demands about 50 percent year over year. GNA needs to do this because it is not efficient to do it locally. As research becomes global, GNA needs to reach out on a global scale. Doing that on an individual basis, is extremely expensive.

Why is GEANT not doing this for Europe, so that GNA can connect with GEANT inside Europe and that GEANT, internationally, connects to the rest of the world, we wanted to know.

René Buch said that GEANT connects a lot of things inside Europe and does an excellent job there but its tradition has not been very strong on the international side. One of the reasons for that is that the different NRENs have different needs. If you are a German university the predominant amount of your users will normally speak German and will live in Germany/Austria, roughly. This will keep a lot of the traffic in the country. The same goes for France, Spain, Italy, and probably Greece, but from the Nordics, being rather small countries, 90 percent of the traffic goes out of the country. As soon as you don't have a national sites, you go to the U.S. or to the UK. This means that the biggest portion in the Nordics, compared to the German site, is out of region traffic, whereas the largest portion of German traffic is inward region traffic. That is one of the big differences. Different NRENs have different perspectives.

The Nordics have a need for global reach because peering and IP transit are important because there are many users who, like students, researchers or technology people, continually are on Amazon, Dropbox or any other service out there. Those services are often placed in conjunction to a fast Internet exchange point. That is one of the reasons why NORDUnet has a very large network. Very early on, NORDUnet saw that this was a large benefit for the Nordic countries.

A large benefit of the good collaboration within GNA is that you don't have 50 different standards for operation nor 50 different procedures of how you handle things with each other. This makes any setting up circuits and any interaction between networks easier than usually before. Otherwise, you have a set of parameters and configurations here and another set of parameters and configurations there and you have to spend 400 e-mails to try and tie things together. GNA has defined the interfaces on both sides. This means that connectivity and setting up service flows is easier between the different NRENs.

There is also a standards organisation about how NRENs are working today so there are many aspects of what GNA is trying to do. It is not just one thing. The collaboration, the power purchasing, and the sharing of resources are the tricky elements.

The kind of traffic that you are doing, is that basically all kinds of research traffic? In the past, you already had some collaborations with optical networks for special applications but you do it now for all the Internet traffic, don't you?

René Buch answered that GNA does it for all Internet traffic as long as it is from research organisations. In the early days, research traffic was limited to two institutions that had both educational purposes together. It doesn't work like that anymore. You have institutions putting their education material on YouTube, some of them are putting it on other sides. You got research people who buy computation power on Amazon Web Services or other Cloud providers. So it is not a clearcut scenario that it needs to be research on one end to be research traffic or educational traffic. The old definition that it can only be research and education when researchers and educators are talking together is so far outdated that it does not make any sense anymore. There are some government agencies that stick to the old definition but it does not really matter, not from NORDUnet's perspective.

So, it is the traffic that the organisations that are part of NORDUnet generate?

René Buch confirmed. He said the only thing that they have is research students so the traffic generated by them is research and education traffic. They watch YouTube videos for several purposes. NORDUnet does not want to judge or play the policeman by saying: 'This is not research traffic'.

If is a bit complicated. You can take the big instruments, like LHC. Between two Tier-1 sides that is where some real research traffic happens, but you would lose 99 percent of the research traffic if that is the only destination. You need to be far more open. What is education traffic? Some universities use in-house or Cloud solutions. The Cloud solutions need to be operational 24/7 because the students are working for exams. This might be in the U.S. or in Ireland, who knows? Things have changed dramatically. Some organisations say that they need to put money only into the high-end research infrastructures but this is only a small portion of the research education work.

Some of these landmark applications are getting more attraction and interest of policy makers and the general public.

René Buch said that the education level of politicians is frightfully low in that sense. They can understand ESFRI projects, ELIXIR, and SKA but if that is the only thing we really care about, GNA would be much smaller. Education and research traffic goes far beyond any of those landmark projects. The problem is that these landmark projects are very good at announcing results. For the politicians it is easier to comprehend CERN or SKA than 30,000 students who want to do different things or 100,000 environmental organisations that do modelling on Amazon Web Services and start sharing data. This becomes way to complicated for politicians mentally to understand.

The GNA is an organisation where different NRENs are working together. How many are involved today?

René Buch replied that it depends on what one means by 'involved'. At the GNA website you have an overview of the technical workgroup consisting of 20 to 25 elements and the number is still increasing. Anyone who wants to contribute - and it is increasing really rapidly - is welcome. GNA is not an exclusive club. If Malaysia wants to contribute to GNA with a link from Singapore to Kuala Lumpur and GNA makes this compliant with the technical specifications, they can participate in the working group and contribute what they want. Then they are part of GNA. You don't need to be a big player in transatlantic or transpacific traffic. Anyone who acts to benefit the community, can be part of GNA.

René Buch believes that GNA can become a really strong driver. One of the hidden things is that by doing this kind of collaborations, the partners come together and the technical staff gets to know each other. This is really beneficial. They can talk about standards. If GNA has 24/7 operations, and some of them are between the American, Australian and Japanese NRENs, it becomes cheaper because monitoring the network during nighttime is fairly trivial. GNA is not going to have more money but it is still going to get fifty percent more traffic year over year. If you do traffic projections like that, it will be really fun in ten years' time. One needs to work smarter and GNA is also part of this smarter working because it is easier to collaborate.

When were these standards and architecture defined?

René Buch answered that it started in 2014 when a group was put together. In the beginning, the group was small but it has increased. There has been a couple of years of intense work. Everybody can get into contact with GNA and associate with it. This is a facilitation to work together. There are two main contacts, namely Erik-Jan Bos and James Williams. GNA has a reference architecture, a specification for the exchange point and a description of how GNA will use MPLS. It may sound trivial but it is not, according to René Buch. There are so many parameters about how you will operate, technical knowhow, and about how you will define things. There is a definition of the Commons on how GNA will actually contribute capacity to the common good. Everybody who participates has to agree on these parameters. That is really a strong point for GNA.

What have you experienced this far. Connecting networks is a nightmare for people but you probably call it a challenge?

René Buch explained that by the things GNA has been doing, the challenge has become smaller because GNA defined interface and the terms. So it should be easier to set things up for new people and make their network GNA-compliant in order to connect with others. With a group of 30 to 35 technical people from all over the planet, GNA has accomplished this. This is a really impressive standard.