Transport networks have witnessed two significant trends over the past half-decade or so. The first has been an explosion in the bandwidth these networks can support and the distances over which they can support it. This is due to the advent of cost-effective wavelength division multiplexing (WDM) and dense-WDM (DWDM), as well as a slew of technologies that extend transmission range, such as sophisticated optical amplifiers. The second has been the need to support a variety of traffic types (voice, video, data) and services: virtual private networks (VPNs), high-speed Internet (HSI), video-on-demand (VoD) and videoconferencing, and IPTV, to name a few. This is due to the need to simplify the network by collapsing intermediate layers and protocol stacks, thus reducing interface and node counts (and, hence, cost) in the carrier network. Thus, transport networks have migrated from being primarily voice-dominated to multi-service supporting infrastructures.
In the past, the optical transport networks themselves did not need to be service- or traffic-aware, as there were a number of layers of multiplexing and aggregation between the carried traffic and the actual transport “pipes.” Indeed a typical protocol-stack layering might take IP data, encapsulate it in Ethernet frames, segment and package those into ATM cells that would be packaged into SONET/SDH frames, which would then ride on an optical wavelength. By contrast, the move today is increasingly towards an optimized stack, which consists of IP data encapsulated in Ethernet frames that (with appropriate framing) ride directly on an optical wavelength — the so-called “optical Ethernet” solution.
Advances in Optical Layer and Network Equipment
So, what are the advances that are making this possible, especially in metro networks? To understand this, we will briefly look both at the optical-layer advances and the network-equipment advances, which constitute some of the keys to optical Ethernet network design.
The fundamental optical layer advances have been the enhancement of WDM technologies with the advent of: Erbium-Doped Fiber Amplifiers (EDFAs), Arrayed Waveguide Gratings (AWGs), and Reconfigurable Optical Add-Drop Multipliexers (or ROADMs). EDFAs enable multiple optical signals, on different wavelengths, to be amplified simultaneously, without requiring expensive conversion into the electronic domain. AWGs, on the other hand, act as an optical filter, and provide a simple mechanism to insert/multiplex and extract/de-multipex optical signals to/from a fiber. In more recent years, ROADMs based on either the wavelength blocker or wavelength selectable switch (WSS) sub-systems have been deployed. These allow any possible wavelength, or a combination of wavelengths, to be added or dropped at a node, thereby allowing providers the flexibility to reconfigure their networks based on traffic needs, leading to true agile optical networks.
In the network-equipment domain, the main advances have been the development of next-generation systems that can support SONET/SDH (TDM data) and IP/Ethernet (packet data).
Legacy networks were built using the TDM paradigm of SONET/SDH, which served as an excellent way to groom voice-dominated traffic and then provision aggregated traffic trunks over the fiber, providing excellent reliability and availability. With the growing dominance of data traffic, SONET/SDH, with its need for synchronization and its limited ability to support flexible bandwidth increments, became increasingly inefficient at meeting the needs of data communications and, hence, a cost barrier. Ethernet, which was already dominant in the LAN, was proposed as a migratory technology, moving to the WAN, in the now quite popular IEEE 802.3z and IEEE 802.3ae standards for 1 Gbps and 10 Gbps speeds, respectively. The less-stringent timing needs of Ethernet made it a lower-cost alternative to SONET/SDH for data services.
In the last year and a half or so, however, there has been rapid rise both in video services (in the form of streaming video, video conferencing, as well as IPTV) as well as in enterprises wanting Ethernet pipes with flexible bandwidths to connect into their WANs. This has posed a technical challenge primarily because traditional Ethernet does not have the deterministic qualities of SONET/SDH in terms of reliability and availability. Extensive work is underfoot in the Metro Ethernet Forum (MEF) to imbue Ethernet with great protection and management (OAM) capabilities, and within the IEEE, in the 802.1Qay WG, to develop Provider-Backbone Bridging-Traffic Engineering, which involves giving Ethernet networks the ability to set up managed, traffic-engineered paths. This is achieved by turning off the MAC learning capability of Ethernet, and, instead, programming (using management) the forwarding tables at every node, thus precisely controlling the path taken by different flows through the network. Measures such as these strengthen the “carrier-class” capabilities of Ethernet.
The simultaneous existence of Ethernet and SONET/SDH services over fiber networks has meant that platforms are now being deployed that cater to a broad mix of these services. These systems, which can cater to a broad range of client-side technologies ranging from Ethernet to SONET/SDH to Fiber Channel and transport these over high-speed WDM networks, are commonly known as Multi-Service Provisioning Platforms (MSPPs). When conjoined with WDM nodes in a single box/rack, they are also known as Multi-Service Transport Platforms (MSTPs).
Need for QoS and Dynamic Provisioning
While ROADMs and MSPPs are ideal platforms to support dynamic configuration of the network to meet traffic needs, there is a need to control, provision, and manage the optical networks in a systematic, automated way. In addition, there is also a strong requirement to meet customer expectations in terms of “Quality of Experience” by deploying mechanisms for establishing and enforcing end-to-end Quality-of-Service (QoS). This has led to the adoption of a suite of protocols, called the Generalized MPLS protocols, developed by the IETF (Internet Engineering Task Force) that is used to control optical networks, both at the physical layer and for traffic management and service provisioning. The GMPLS suite of protocols includes routing protocols that are used to discover network topology and available resources (bandwidth, timeslots, and wavelengths) in the optical Ethernet transport network, and signaling protocols that are used to signal the setting up of active (and backup) paths through the network. The GMPLS protocols can set up “paths” that are comprised of a sequence of wavelengths, time-slots, or packets/frames that share a common characteristic (such as being headed to the same destination or belonging to the same Class-of-Service).
From a cost perspective, the primary cost is in the electronic packet engine that aggregates multiple lower-rate signals into a single high-speed electronic signal. Typically this high-speed electronic signal is then translated into an ITU-grid optical frequency (wavelength) by a sub-system called the transponder. A critical timing and cost-optimization challenge is in placing multiple lower-rate signals as client interfaces in the same subsystem that also houses the ITU-side optics. Integration of the electronic packet multiplexer (using Ethernet technologies) with the optical transponder constitute the main challenge for providers in meeting the paradigm of dynamic bandwidth provisioning, especially for their small- and mid-sized customers that dominate much of metro core and metro access/collector markets.
Evolution at the Optical Layer and Ethernet Layer
ROADMs form the central feature of metro optical networks, the largest business case for optical transport, with three generations of architecture: the Fixed OADM (FOADMs), the Reconfigurable OADM (ROADMs, contemporary) and the Dynamic OADM (DOADM). FOADMs allow dropping and adding of wavelengths at a node with the constraint that only a fixed set of wavelengths can be dropped (limiting dynamism) and the ports from which the wavelengths are dropped (or added) are also fixed. In ROADMs there is flexibility in terms of which wavelengths can be dropped and which can pass-through (called optical bypass), but there is a restriction on the mapping between wavelengths and ports. The DOADM is considered the ultimate in terms of flexibility, and allows dropping/adding of any wavelength at any port in a node thus allowing full flexibility in the network and thereby reducing operational expenditure (e.g. maintaining a smaller inventory of transponders).
At the same time, developments in Ethernet OAM standards (e.g. IEEE 802.1ag and IEEE 802.3ah) allow for performance monitoring and maintenance of end-to-end and Ethernet circuits and each Ethernet hop, respectively, by keeping track of parameters such as transmitted/dropped frames, frame delay, jitter and loss, and availability. These allow operators to perform diagnostics, manage their networks, and deliver service assurance.
Looking to the Future
Optical transport is poised to enter a new age. The rise of carrier-class Ethernet along with an acute need for bandwidth intensive services (such as video streaming) implies that future optical networks must be able to dynamically allocate bandwidth (on-demand) to nodes, support good optical-layer multicasting, and provide for lower-cost solutions that can be implemented in smaller networks in the metro access as well as enterprise markets. This growth has forced research in higher-speed solutions — such as 40 Gbps SONET/SDH and 100 Gbps Ethernet (100GigE). Both of these technologies have significant physical layer issues and impairments such as modulation format, timing issues (e.g. pulse width of 10 picoseconds in 100GigE), dispersion compensation, and OSNR monitoring.
In the metro, the dynamism requires that newer solutions would have to use high-speed algorithms for bandwidth provisioning as well as architectures that can support dynamic allocations. There are principally three schools of thought emerging for design of metro optical Ethernet transport:
- fully electronic grooming solutions, such as all-Ethernet packet transport and the Provider Backbone Bridging — Traffic Engineering (PBB-TE), IEEE 802.1Qay, initiative;
- all-optical grooming solutions, using interleaved access such as burst switching or wavelength buses called light-trails;
- digital optical networks using photonic integrated circuits (PICs) based on Indium Phosphide technology.
Approach (1) is evolutionary from a technology perspective but has several capital requirements at high-speeds. Approach (2) is something of a paradigm shift and can be done in incremental steps, with partial electronics and partial optics, e.g. light-trails. Finally, approach (3) is somewhat revolutionary and requires complete revamping of existing optical networks, but it does have the potential to deploy the System-On-Chip (SOC) concept, thereby drastically reducing growth costs.