Comparing Optical & OTN Switching
Architectures in Next-Gen 100Gb/s Networks
Serge Melle, Abhijeet Deore, Onur Turkcu,
Satyajeet Ahuja, and Steven J. Hand
Infinera Corp, 140 Caspian Court, Sunnyvale, California, 94089, USA
smelle@infinera.com
Abstract: Bandwidth efficiency for 100Gb/s WDM and switch architectures are
compared in a North American long-haul network. Results show that OTN
grooming utilizes fewest WDM wavelengths, and integrated WDM/switch
architecture utilizes fewest client service interfaces.
© 2012 Optical Society of America
OCIC Codes: : (060.0060) Fiber optics and optical communications
1. Introduction
As optical transport networks evolve from 10Gb/s to 100Gb/s wavelengths, current research suggests that service
data rates remain primarily at 10Gb/s or less. In 2011, over 98% of services in long-haul networks 10Gb/s or less,
with >10Gb/s services forecast to grow to only 5% of total interfaces by 2017 [Ref 1]. Past studies have shown that
use of muxponders led to an inefficient use of bandwidth [Ref 2]. Thus a key focus of investigation in this paper is
to determine the usefulness of employing multi-Terabit, OTN switches in next-generation 100Gb/s WDM networks,
and whether transponder/ROADM architectures without OTN switching remain appropriate, efficient and cost-
efficient. We compare network architectures that utilize WDM/ROADMs with or without OTN switches, and
present comparative results for WDM and client ports, reflecting relative CapEx costs of each architecture.
2. WDM and OTN Switching Architectures
WDM Only: This architecture utilizes trans/mux-ponders to map services onto a 100Gb/s WDM interfaces. When
the service data rate approximates the WDM data rate, for example 100 Gigabit Ethernet, it is mapped using a
transponder. When the service data rates are less than the WDM data rate, for example 10/40Gb/s, this is done using
a muxponder with either 10x10Gb/s or 2x40Gb/s client ports. Services are mapped at the edges of the network and
routed all-optically through intermediate ROADM nodes as far as possible, as shown in Figure 1. This provides a
point-point, end-to-end wavelength connection across the network, with no intermediate optical-electrical-optical
(OEO) conversion, unless regeneration of the signal is required. When regeneration was required, manual grooming
at regen points was done for services transiting a common wavelength path, to enable a modicum of traffic
grooming, albeit at the cost of manual interventions and consequent OpEx costs.
WDM plus Stand-Alone OTN Switch: This architecture utilizes the same trans/mux-ponder architecture described
above, and adds the use of a digital OTN switch at Tier 1 and 2 add/drop nodes, as shown in Figure 2 (a). The
stand-alone OTN switch is deployed separate from the WDM systems, and connects using short-reach optics, which
can drive the use of many back-back fiber inter-connections between the WDM and OTN systems.
Integrated WDM/Switch: This architecture assumes a OTN switch with integrated WDM interfaces at all network
nodes, as shown in Figure 2 (b). The convergence of the WDM and OTN into a single system eliminates use of
short-reach inter-connections between separate WDM and OTN systems and reduces space and power use.
Both architectures with OTN switching enable intra- and inter-wavelength grooming, allowing sub-wavelength
services to be multiplexed and reduce use of WDM ports. In all three cases, WDM designs assumed an optical
reach for 100Gb/s wavelengths of 1500km, with the ability to support optical express pass-through up to 12
intermediate ROADM nodes without impacting reach. If the end-to-end optical path exceeded the reach of the
wavelength, back-back WDM optics were used for regeneration. If the payload of a wavelength is fully utilized,
either because it is carrying a 100Gb/s service demand, or a full complement of sub-wavelength services, then that
wavelength is routed as far as possible, making maximum use of optical bypass at intermediate nodes.