ExpertiseMulti layer network architecture

Multi layer network architecture

Network architecture visions

How should we build future open access network to meet the open access business model? How does network virtualization and the combination of electrical and optical forwarding lead to higher the network energy efficiency? How do we balance the trends of access node consolidation and highly local traffic patterns due to locality aware content distribution networks?

In a world where convergence is the motto, where networks only become more extensive and complex, and where environmental effects are becoming more important, there are many new technical, operational and financial challenges. We offer for example:

  • Research and development as support in the internal development process, such as analysis and design for the migration to IPv6, future-proof networks and network services.
  • Fast prototyping and emulation of new network functions, such as how power management algorithms affect the network and its network elements.
  • Support in standardization such as IETF.

Software Defined Networks – a new standards is evolving

Learn about Open Flow Together with five other partners, RISE Acreo is working on the emerging topic of Software Defined Networking (SDN), primarily within the framework of the EU project SPARC. Sprung from the US GENI project and Stanford University, SDN looks like a promising way forward when it comes to network control and management. Additionally, it should improve network flexibility as well as enabling faster innovation in network functionality.

In the traditional router model seen in figure 1) a router’s control plane communicates with the outside world using various protocols and then autonomously decides what actions to take. Typically this involves a process running within a closed operating system calling a proprietary API which in turn causes the operating system to program specialized forwarding hardware, again using a proprietary API. Adding new functionality to a routerusually involves standardizing a new protocol that (often) re-invents distribution-, signaling- and other mechanisms, and then waiting for vendors to implement the new protocol. Comparing this process to how it works in the PC world reveals why network development is moving so slow compared to other areas. In the PC world, where well defined and open APIs exists, anyone can quickly add functionality by simply writing a piece of software and installing it on their machine, often reusing existing functionality through software libraries.

SDN proposes to break apart the traditional model by creating open APIs between the hardware and the operating system,and between operating system and network applications. In the SDN model (seen in fi gure 1) a Network Operating System (NOS) is responsible for maintaining an up-to-date view1of the network and its current state, the NOS may be implemented as a distributed system running between several servers in order to increase its resiliency and scalability. The NOS does not only maintain a view of the network but is also responsible for accepting changes to the view and implementing the changes on the network hardware. Changes to the view come from network applications running on top of the operating system, these are software modules that are able to access the network view maintained by the NOS as well as modifying it. Adding new functionality becomes much simpler; just write a software module utilizing the API provided by the NOS and the NOS is responsible for updating the network and distributing the new state.

Most well known in the SDN world is OpenFlow, an open protocol designed to expose the internals of a router or switch and provide functionality to modify it. The protocol models a routers internals as a FlowTable, a table containing rules that can be used to match an incoming packet, for example destination IPv4 address 1.2.3.4 with TCP destination port 80, and a number of actions, for example, modify the destination address and output the packet on interface 4. If an incoming packet does not match an existing rule in the router, the packet can be sent to the NOS where a network application can investigate further, and perhaps install a new rule that takes care of packets belonging to this particular fl ow of packets. The OpenFlow protocol is currently in heavy development and is by no means complete, it is not unlikely that updated protocols will appear in the future once it has become clearer how one should actually see and interact with the internals of a router.
While the field of SDN is still in its infancy, an official standardization body (the Open Networking Foundation2) was established just last year, shifting the control from academia to industry. Several products are already available, for example from NEC and IBM. These products typically focus on datacenter applications where complicated network setups are diffi cult to manage using existing protocols. In the SPARC project we have chosen to focus on a different area, namely ISP access- and aggregation networks. In this project we are investigating if it’s possible to apply the SDN approach to simplify carrier networks, with all the requirements on reliability and resiliency that comes along with that.

In one of the SPARC demonstrations, can be seen in figure 2), we have integrated an SDN controlled OpenFlow-MPLS network with an existing legacy IP/MPLS controlled core network. The routers in the aggregation network only “speak” OpenFlow to the NOS. The NOS in turn is responsible for communicating through OSPF, LDP, RSVP-TE and BGP to the legacy network. Information reaching the NOS from the legacy network triggers it to calculate paths through the OpenFlow network and establish point-to-point and multipoint-to-point connections for services, for example Internet connectivity or IPTV transmissions. If links in the OpenFlow network fails, either the NOS quickly calculate new paths and re-establishes connectivity, or the switches themselves fail-over to pre-established protection paths. While the SPARC project itself is ending in 2012, we are continuing with further research within the promising and fast moving area of SDN.

Pontus Sköldström, Acreo
pontus.skoldstrom [at] acreo.se

People

Anders Gavler
Group manager - Applied Networking
+46 70 652 49 38
anders.gavler [at] ri.se

Stéphane Junique
Senior Scientist
+46 (0)70 772 77 48
stephane.junique [at] ri.se

Publications