Traffic engineering and topology design considering multilayer configuration have become more important. While multilayer design studies usually discuss the traffic engineering issue or reliability, this paper focuses on network construction cost in studying multilayer topology design. The number of ports for the IP layer and the maximum number of Wavelength Division Multiplexers (WDM) for the optical layer are used as construction cost factors. Given a traffic matrix for the IP layer, 1) the number of ports is minimized to obtain a topology and a traffic matrix for the optical link, and 2) the maximum number of WDMs is minimized to configure the optical layer topology. It is shown that both the IP and Optical path layers have been given economic topologies. We present LP formulations of this scheme and the results of a simulation of the full-mesh traffic of 5 nodes, which shows that both layers are successfully optimized.
The expectations placed on the IP optical network have grown with the increase in new network applications intended for broadband services, such as IP phones and stream broadcasting. It yields large-capacity and economic networks that offer IP technology for packet-bypacket routing and optical cross-connect technology for optical switching. Because this network essentially has a multilayer configuration, careful design of both layers is needed for topology establishment and traffic accommodation. While backup path reliability problems [1-3] and traffic engineering [4-6] for traffic smoothing have been actively studied recently in terms of both layers [7-9], it is seems that topologies are usually predetermined and fixed.
This paper discusses the design of both topologies. It is intended to identify the most economic topology configuration that can also accommodate the traffic. This paper is organized as follows. Section 2 introduces the problem of multilayer topology design considered here. Section 3 provides formulations and the calculation procedure to solve it. Sections 4 and 5 verify its effectiveness by a simulation that uses an example traffic matrix. Finally, we conclude this study in Section 6.
An IP-optical network is comprised of the IP layer (logical layer) with its logical links and the optical path layer (optical layer) in which optical links are formed. Logical links between nodes in the logical layer are defined by the interfaces of the end nodes. An optical path in the optical layer realizes a logical link as a physical entity to provide its bandwidth. Each logical link uses an optical path in the optical layer. Our goal is to design economic topologies of the logical and optical layers when full mesh traffic demands between each node are given.
The following conditions are assumed here:
1) Logical layer: The number of interfaces determines the cost of the logical layer, so the total number of interfaces should be minimized. We assume that only one type of interface, such as 10 Gbps, is used.
2) Optical layer: The number of WDMs determines the cost of the optical layer, so the maximum number of wavelengths should be minimized. Each node pair linked by one optical fiber or is not linked. A WDM is used when one fiber carries more than one optical paths. Here, transparency of optical path through wavelength continuity is not considered. In other words, cross-connects with OE/EO functionality are available. Optical paths have the same bandwidth regardless of wavelength.
Figures 1 and 2 provide an example of routing of traffic and logical/optical links, respectively. Routing for two traffic demands (traffic 1, and traffic 2) in the IP layer is shown in
here that flow splitting and merger can be conducted in a route such as traffic 1.
For the logical layer, the total number of ports for all nodes is used as the cost factor. It is assumed that the total cost depends on the number of ports while the installation of nodes is essential in the logical layer. For the optical layer, the maximum number of WDMs is important for cost effectiveness. It is considered that the number of wavelength paths is suitable even for the case in which the wavelength paths are implemented over existing facilities. Even in this case, it is assumed that the maximum number of WDMs is used as the cost measure since WDM device cost increases with its scale.
The design procedure in this paper has two steps:
1) Logical links with the fewest (total) interfaces are determined from nodes of the logical layer, traffic between each pair of nodes, and the bandwidth of each interface. The link capacity necessary between each pair of nodes is determined.
2) The capacity obtained in the above step is used as the input traffic, and the topology that minimizes the maximum number of WDMs between each pair of nodes is determined.