The growth of the size of the routing tables limits the scalability of the conventional IP routing. As scalable routing schemes for large-scale networks are highly demanded, this paper proposes and evaluates an efficient geometric routing scheme and related low-cost node design applicable to large-scale networks. The approach guarantees that greedy forwarding on derived coordinates will result in successful packet delivery to every destination in the network by relying on coordinates deduced from a spanning tree of the network. The efficiency of the proposed scheme is measured in terms of routing quality (stretch) and size of the coordinates. The cost of the proposed router is quantified in terms of area complexity of the hardware design and all the evaluations involve comparison with a state-of-the-art approach with virtual coordinates in the hyperbolic plane. Extensive simulations assess the proposal in large topologies consisting of up to 100K nodes. Experiments show that the scheme has stretch properties comparable to geometric routing in the hyperbolic plane, while enabling a more efficient hardware design, and scaling considerably better in terms of storage requirements for coordinate representation. These attractive properties make the scheme promising for routing in large networks.

In OpenFlow, data and control plane are decoupled from switches or routers. While the data plane resides in the switches or routers, the control plane might be moved into one or more external servers (controllers). In this article, we propose verification mechanisms for the data plane functionality of switches. The latter consists of two parts: (1) Flow-Match Header part (to match a flow of incoming packets) and (2) action part (e.g., to forward incoming packets to an outgoing port). We propose a mechanism to verify the Flow-Match Header part of the data plane. The mechanism can be executed at the controller, or on an additional device or server (or virtual machines) attached to the network. Deploying a virtual machine (VM) or server for verification may decrease the load of the controller and/or consumed bandwidth between the controller and a switch. We propose a heuristic to place external verification devices or VMs in a network such that the verification time can be minimized. Verification time with respect to consumed resources are evaluated through emulation experiments. Results confirm that the verification time using the proposed heuristic is indeed shortened significantly, while requiring low bandwidth resources.

The first global interoperability experiment of GMPLS controlled Ethernet with VLAN tag swapping between two different implementations is successfully demonstrated. High definition video streaming is realized through a newly established Layer 2 Label Switched Path (L2-LSP). The results of this experiment can be applied to designing reliable Layer 2 networks.