2476 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 21, NO. 11, NOVEMBER 2003 Waveband Grooming and IP Aggregation in Optical Networks Rajendran Parthiban, Rodney S. Tucker, Fellow, IEEE, and Chris Leckie Abstract—An automatically switched optical network (ASON) can be used as the transport layer of generalized multiprotocol label switching (GMPLS) networks. The design of an ASON in- volves determining the number of optical cross-connects (OXC) in the network, the required number of ports per OXC, and the in- terconnection topology of the OXCs. Given the number of ports per OXC, we present a linear algorithm to find the number of OXCs and to identify a cost-effective topology. We then develop a scheme that can be used to perform waveband grooming for several different topologies of an ASON that uses single-layer multigran- ular OXCs. We identify the bottlenecks and investigate the effect of traffic grooming schemes in the design of an ASON as a function of the peak access rate per customer. We evaluate the topologies and architectures for a national trunk network. Index Terms—Automatically switched optical network (ASON), generalized multiprotocol label switching (GMPLS), IP/optical aggregation, waveband grooming. I. INTRODUCTION G ENERALIZED multiprotocol label switching (GMPLS) networks supported by label switching routers (LSR) are favored for the backbone IP network because of their ability to provide faster switching at the core of the network and to support different levels of quality of service (QoS). In GMPLS, the traffic between the LSRs can be sent in the form of lightpaths [1]. In an automatically switched optical network (ASON), lightpath connections are established with optical cross-connects (OXC) which provide fast and flexible reconfiguration of lightpaths [2]. Hence, ASONs can be used as the transport layer for the GMPLS network [3]. The traffic to an ASON can be aggregated at different levels—packets aggregated into a time-division multiplexed (TDM) frame, TDM frames aggregated into a lightpath, lightpaths aggregated into a waveband, and wavebands aggregated into a fiber [4]. The aim of this paper is to analyze how we can use different network topologies and traffic aggregation schemes to design the ASON transport layer of a GMPLS network. The design of an ASON involves finding the number of OXCs, the required number of ports per OXC and the intercon- nection topology of the OXCs. Unlike the design problems of Manuscript received November 20, 2002; revised July 16, 2003. This work was supported by the Australian Research Council. R. Parthiban and R. S. Tucker are with the ARC Special Research Centre for Ultra-Broadband Information Networks, University of Melbourne, Australia (e-mail: r.parthiban@ee.mu.oz.au; r.tucker@ee.mu.oz.au). C. Leckie is with the ARC Special Research Centre for Ultra-Broad- band Information Networks, University of Melbourne, Australia (e-mail: c.leckie@ee.mu.oz.au) and also with the Department of Computer Science and Software Engineering, University of Melbourne, Australia. Digital Object Identifier 10.1109/JLT.2003.819136 traditional telecommunication networks, the design problem of an ASON should include the wavelength continuity con- straint [5]. In solving the problem, some researchers [6], [7] start with a fixed number of OXCs and determine the interconnection topology. This problem is NP-hard and needs heuristic algorithms. In [6] and [7], the problem is solved by using the concepts of the traditional telecommunication design algorithms such as the Saturated Cut Algorithm [8] and MENTOR [9], and modifying them to include the wavelength continuity constraint. Other researchers [10] tackle an even harder problem in which the number of ports per OXC is given and the number of OXCs and interconnection topology are found. In [10], a stochastic optimization approach is used, which is computationally complex. It is assumed in [10] that the node (OXC) degree is equal to the number of ports per OXC. In other words, it is assumed that each OXC is capable of reconfiguring links. However, the effects of aggregating the lightpaths into wavebands, fibers or links are not considered. We simplify the problem of finding the number of OXCs and the network topology by starting with symmetric topologies. We find analytically tractable solutions for symmetric topologies and hence reduce the computational cost. This enables us to find the number of nodes (OXCs) and the topology of the optical network through a linear algorithm. The topology is then optimized using the geographical distribution of the customers to remove the symmetry requirement. We also investigate the effect of traffic aggregation schemes—IP aggregation and waveband grooming—on the topology and the overall network cost. Traffic grooming at the IP layer, also known as IP aggrega- tion, can be used to reduce the traffic load in the optical layer [11]. In particular, IP routers can be used to aggregate mul- tiple traffic flows onto a single wavelength and hence reduce the overall number of wavelengths used in the network. Harai et al. [12] have devised an algorithm to decrease the overall number of wavelengths of an optical network when one knows the place- ment of IP routers. In this paper, we place the routers at the edge of the optical network and investigate how this affects the overall network cost. Traffic grooming can be performed at the optical layer by grouping lightpaths into a waveband and wavebands into a fiber. Ordinary OXCs allow lightpath to be switched as a single entity (i.e., using a single port). Multigranular OXCs (MG-OXCs) are used to switch traffic at multiple levels of granularity, namely, at the fiber, waveband or lightpath levels. In MG-OXCs, each level of granularity (i.e., lightpath, wave- band, or fiber) is switched using a single port [13]. Two types of MG-OXC—single-layer MG-OXC [14] and multi-layer 0733-8724/03$17.00 © 2003 IEEE Authorized licensed use limited to: Georgia State University. Downloaded on June 28, 2009 at 19:17 from IEEE Xplore. Restrictions apply.