ICTON 2012 We.B3.1 978-1-4673-2227-0/12/$31.00 ©2012 IEEE 1 Artificial Bee Colony Model for Survivable DWDM Optical Networks Design Arash Rashedi*, Yousef S. Kavian**, Ali Mahani*** *Iran Telecom Company, Khouzestan, Ahvaz, Iran ** Faculty of Engineering, Shahid Chamran University, Ahvaz, Iran *** Faculty of Engineering and Technology, Shahid Bahonar University, Kerman, Iran e-mail: y.s.kavian@scu.ac.ir ABSTRACT This paper presents an application of Artificial Bee Colony (ABC) algorithm for survivable optical network design. The dedicated path protection (DPP) survivability scheme is modelled by ABC algorithm for establishing two link-disjoint paths between each node pair in demand matrix considering single link failure scenarios. In the proposed algorithm every food source represents a lightpath between each node pair in optical network and the position of food sources are modified by artificial bees to minimize the total number of utilized wavelengths by working and spare lightpaths. Simulation results demonstrate that the proposed ABC based survivability scheme is able and efficient to design fault-tolerant optical transport networks. Keywords: optical networks, intelligent algorithms, survivability, artificial bee colony, dedicated path protection, single link failure. 1. INTRODUCTION Reliable communication services and systems are mainly demanded by end users for truly information exchange in web-based applications. Therefore, the ability of communication systems and networks to work and continue service providing at acceptable level in the presence of attacks and failures is an important designing issues. Network layers employ different policies and algorithms to deal with about errors, failures and attacks causing by different sources. The higher layers are more time and resource consuming comparing to lower layers. For example the optical layer provides resource and time efficient fault-tolerance schemes rather than upper unprotected layers in IP-over-DWDM networks. In reliable optical backbones, the working lightpaths carry data and information under normal and fault-free conditions and the spare lightpaths which are link-disjoint with working ones carries data after failure events. Since the primary lightpaths and corresponding backup lightpaths are link-disjoint, there is a valid lightpath after any single link failure scenario. The protection based algorithms pre-compute and reserve backup resources in advance during network connection setup and planning which are used for failure recovery applications. Therefore, affected traffic will be rerouted through backup paths after failure occurrence. In terms of resource allocation paradigm, protection architecture can be applied via dedicated and shared resource utilisation. Dedicated protection devotes backup resources for the primary connection, unavailable to any other entity. Dedicated protection provides greater connection availability than shared protection. Shared protection is more efficient than dedicated protection for capacity trading but as the degree of sharing increases, the connection availability is reduced and longer recovery time is required than dedicated protection and backup resources need to be configured after failure occurrence. Protection can also be considered as link protection and path protection. In link protection, a link- disjoint backup lightpath is reserved for each individual link in working lightpath. Path protection refers to backup lightpath reservation for each working lightpath. Path protection usually has lower resource requirements and lower end-to-end propagation delay than link protection. This paper presents an application of ABC optimization algorithm [1, 2] which is an extension of ABC based RWA model [3] to design fault tolerant optical transport networks by establishing a pair of working and spare lightpaths for each connection request in a demand matrix using the dedicated path protection (DPP) architecture. 2. PROBLEM STATEMENT The optical network is represented by a connected graph G(N, L) where N is the set of nodes and L is the set of connecting links. All optical links are assumed to be bi-directional and the total number of wavelengths that can be carried on each link is bounded by W. A typical connection request R in optical network represents a source node (a), a destination node (b) and number of requested wavelengths (w). Minimizing total network planning cost considering a linear capacity cost function is proposed where the link cost is the cost of allocated wavelengths to working and spare lightpaths. ( ) l l l L Minimize A B ∈ + ∑ , (1)