Open LightPath Exchanges

http://sne.science.uva.nl/ http://uva.nl/
Introduction Links Related Projects Related publications Related talks


Open LightPath Exchanges, sometimes reffered to with the acronym GOLE's (G for GLIF) play an important role in the worldwide Lightpath topology. At these exchanges LightPath carriers from different parties come together. The exchanges increase dramatically the degrees of freedom for the topology.

The exchanges connect different networks from different (or same) domains. Different networks may use different technologies to forward data through their infrastructure. It is essential to consider the attributes of the data that is used in deciding the forwarding of the traffic. For example the dominant technology for the Internet is based on routers that forward based on the IP addresses. In ethernet the decisive attributes are typically source and destination mac-layer addresses and vlan-id's. In openflow these are flow-labels and similar attributes are used in MPLS. In a pure photonic network as deployed by SURFnet and I2 wavelength selective switches forwarding is based only on colors. Those devices are completely unaware of labels and addresses. If one wants to set up end to end dedicated lightpaths connections through a generalized dynamic infrastructure consisting of networks and exchanges one has to take the multi technology nature of networks into account.

Given technology cycles of just a few years one has to assume that different technologies meet at exchanges. One approach is to translate everything to a common layer and technology at those exchanges but this would cost degrees of freedom. Another approach is to design exchanges to interconnect on different layers and to translate from layer/technology to layer/technology where appropriate. As such an exchange could serve cross border alien light at the same time as OpenFlow and ethernet vlan switching, whatever the application needs.

In order to make such a system work advanced information about the networks is needed and novel path computation is required. This is the basis of the NDL work pioneered by the SNE group at the University of Amsterdam. Together with SURFnet the UVA developed the concept of Alien Light and demonstrated that at TNC 2011 and TNC 2012.

As many may still remember my research group @ UVA did a study on this topic back in 2004-2005 to understand what the term "open" in the context of different types of (Inter)net exchanges would mean and what kind of implications it would have technology and authorization wise. At this moment my group leads a study of trust models very relevant also in this context of open exchanges.

GLIF_5-11_World_4k.jpg GLIF_5-11_World_4k.jpg
Fig 1. GLIFaround 2012 with the differentGOLE's visible. Fig 2. Networks of ScienceDMZ's and SDX's allowing advanced functionality and security in the networks.

Now in 2017 we study the effects of Software Defined Networking and Function Virtualisation. ESNET developed the notion of scienceDMZ's as a way to structure campus networks such that the high performance network based applications are not hindered by the security measures to keep campus networks safe and that normal campus traffic is not drowning in the high speed data intensive science load. These new technologies give tremendous new opportunities to create new trusted functions in the Internet, not only at the home organisation but also in exchanges and networks upstream in the traffic paths. This is depicted in fig 2.

Related projects

Related publications & talks

Here are some pointers of the work we (mostly Freek Dijkstra and Leon Gommans) did.

Related talks


home home