A comparison of next-generation IP-centric transport architectures

The popularity of the Internet and Internet protocol (IP)-based intranets is promising enormous growth in data traffic originating from IP endpoints, prompting network operators to reconsider network architectures so that they can most effectively absorb the projected growth. At the same time, new technologies are being introduced at a phenomenal pace, providing network operators with numerous and complex choices involving dense wavelength division multiplexing (DWDM), synchronous optical networks (SONETs), packet over SONET (POS), packet over wavelength (POW), and asynchronous transfer mode (ATM). In this paper, we evaluate alternative transport architectures for carrying IP-based traffic using the projected traffic data, nodal configuration, and optical fiber connectivity of a realistic, national-scale IP backbone. We compare the option of carrying IP directly versus IP over ATM for three types of transport architecture: SONET bidirectional line-switched rings (BLSRs); mesh networks of optical (or electrical) cross connects; and DWDMs without underlying optical cross connects (OXCs) - that is, with one or more wavelength links between each pair of IP switches. These options also include restoration choices. SONET BLSRs provide fast restoration based on self-healing ring technology. OXCs provide fast restoration for underlying mesh at the wavelength level. For point-to-point wavelength links, we consider service-level (IP and ATM) restoration. We compare these options in terms of many network characteristics - port counts, circuit miles, wavelength miles, fiber miles, and overall cost - and consider all the critical constraints and flexibilities for each choice.