ICTON 2006 101 Tu.B3.1 ⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯ The work reported in this paper was supported in part by the German Ministry of Education and Research (BMBF) within the EIBONE (Efficient Integrated Backbone) project under contract number 01BP567. 1-4244-0236-0/06/$20.00 ©2006 IEEE Cost-Efficient Transparent Optical Networks with High Connection Availabilities Arie M.C.A. Koster Zuse Institute Berlin (ZIB), Takustr. 7, D-14195, Berlin, Germany E-mail: koster@zib.de ABSTRACT Transparent optical networks have to be designed cost-efficiently in order to survive in the highly competitive telecommunication network market. In this talk we discuss the power of mathematical optimization techniques to obtain such designs. One aspect that deserves particular attention in transparent optical networks is the availability of connections. Several survivability concepts have been developed over the years. We compare the concepts 1+1 dedicated path protection, shared path protection and the recently introduced concept Demand- wise Shared Protection (DSP) on the network cost and connection availabilities. The results show that DSP obtains nearly the same network availability against significantly less cost. Keywords: optical network design, survivability, availability, demand-wise shared protection. 1. INTRODUCTION Transparent optical networks will build the core of next generation telecommunication networks. As such backbone networks connect larger geographical regions, they should be capable to transport high data rates between the network nodes simultaneously. On the hand, the competition in the telecommunication market requires that the network is designed in a cost-efficient way. On the other hand, to satisfy customer satisfaction network operators have to offer connections with high availability guarantee. To protect traffic against node and link failures, many survivability concepts have been proposed in the literature. Their applicability depends on the used technology (e.g., MPLS, ATM, SDH,WDM), and each of the concepts has its particular strengths and weaknesses with respect to investment cost, management effort, and recovery performance. Dedicated protection concepts (like 1+1 or m : n dedicated path protection) are easy to implement in terms of network management and provide short recovery times, but the resulting networks are often rather expensive. In contrast, shared protection concepts (like m : n shared path protection) make more efficient use of backup capacities and thus can potentially reduce network cost, at the expense of a more complex network management. Recently, the concept of Demand-wise Shared Protection (DSP) has been introduced as a novel survivability concept that shares capacity within the demand (i.e., the bandwidth required between a pair of nodes) but does not share it with other demands, see [1][8]. In this way DSP combines the main advantages of dedicated and shared protection. It can be expected to be less expensive than dedicated path protection and, at the same time, easier to realize than shared path protection. In this paper, we compare the concepts 1+1 dedicated path protection, DSP, and shared path protection on their network cost and their connection availabilities. We first discuss a mathematical optimization model that allows to find cost-efficient designs of transparent optical networks with demand routings according either one of the survivability concepts. Next, we discuss how the network availability for such designs can be computed. We conclude with a computational evaluation of the three compared concepts. 2. NETWORK OPTIMIZATION MODEL The cost-minimal network design problem in transparent optical networks is modelled with a (mixed-) integer programming model. This model includes all decisions concerning topology, hardware configuration (e.g., WDM systems, optical cross connects), lightpath configuration (logical topology), and lightpath routing, see [6] and [9] for details. This model can be adapted to either model 1+1 dedicated path protection, shared path protection, or demand-wise shared protection. This network design problem can be decomposed into two major blocks: one describing the hardware configuration, and the other one describing the routing constraints. In this section, we restrict ourselves to the mathematical model for a survivable routing using DSP. The connection between the routing and hardware configuration is established by the decisions on the capacities y e for each link e ∈ E of the underlying (undirected) network graph G = (V,E). These capacities must suffice to accommodate a survivable routing using DSP. A DSP routing has to fulfil the two basic requirements a network operator has: • a pre-determined demand value has to be satisfied in the failure-free network state, and • in any considered failure state, a specified fraction of the demand must survive. Authorized licensed use limited to: RWTH AACHEN. Downloaded on May 26, 2009 at 11:09 from IEEE Xplore. Restrictions apply.