Combining GFP and WDM to deliver transparent MAN services efficiently
Service providers want new levels of efficiency and optimization in metro networks. The fulfillment of this desire faces many challenges due to the wide range of protocols, data rates, traffic patterns, and physical topologies typical of these networks.
In this context, SONET/SDH networks provide multiplexing efficiency, reliability, and timely transport of client signals. However, they present a high infrastructure cost incurred at every site and increased protocol complexity for all client signals. SONET/SDH's high infrastructure costs are caused by its requirement to terminate two large trunk interfaces, the need for redundant STS-1/STM-1 switch fabrics capable of handling the trunk interfaces at every site, and the complex protocol adaptation of all client signals to the rigid requirements of the SONET/ SDH architecture.
The cost disadvantages of SONET/SDH are further magnified as the data rate of client signals approaches the capacity of the trunk interfaces. For these high-speed data services, the benefits of multiplexing efficiency using fine STS-1/STM-1 timeslot granularity is now an underutilized cost burden incurred at every site on the ring. Further, SONET/SDH's built-in reliability through its ring architecture means there is 100% overbuild on every service, even when protection is not required.
The high costs of deploying high-speed data services over a SONET/SDH infrastructure has forced many service operators to bypass their existing network and create parallel "special" networks. Examples of these special networks include dedicated Gigabit Ethernet (GbE) and SANs on separate infrastructures. But this practice is suitable only when a few of these networks will be deployed, when there is an abundance of fiber, when the customer is willing to wait for the separate network to be built, and when the customer can pay the very high price for the dedicated network.
A transparent WDM architecture provides a partial solution because it creates an infrastructure that supports all existing protocols at any bit rate, with the flexibility to support future optical interfaces and applications. This approach is very efficient for high-speed data signals such as GbE, Gigabit Fibre Channel (FC), and OC-12/STM-4 because it eliminates the overhead of protocol adaptation layers and unnecessary signal regenerations.
However, a drawback to purely transparent WDM networking is that a dedicated wavelength is required for all services. That can lead to underutilization of the wavelength capacity for lower-speed optical signals. A means to perform subwavelength multiplexing while preserving the protocol and service transparency is needed to maximize the efficiency of optical networks and lower the cost of offering services.
WDM optical-networking devices such as an optical add/drop switch (OADX) enable wavelengths to be managed in a manner similar to how timeslots are managed in a SONET/SDH network. With WDM, the minimum transport granularity is an optical wavelength. The optical-networking devices can transparently switch wavelengths between the add/drop ports and trunk ports to dynamically add or drop a wavelength to a client interface or to express any wavelength among a set of trunk interfaces. Essentially, the optical-networking device supports all the networking features typically found in a SONET/SDH networking device. These features include client signal multiplexing, protection, bridge-and-roll, and switching.
The optical-networking infrastructure enables a service operator to implement special networks quickly and in a scalable manner. Multiple client networks with diverse protocols and features share a common optical transport network with existing SONET/SDH networks. The special networks can be included in the portfolio of service offerings available to all customers, just like any circuit service over SONET/SDH.
The optical wavelength networking advantages of data protocol and rate transparency are ideal for high-speed services—but for lower-speed services, there can be significant inefficiencies. For instance, OC-3/STM-1 and 100-Mbit/sec Ethernet services are expected to be high-growth areas in the data services market. In fact, the forecasted volumes could exhaust the available wavelengths on a single fiber pair. However, a wavelength operating at 2.5 Gbits/sec has enough capacity to contain 16 OC-3s or 20 100-Mbit/sec Ethernet signals. Therefore, the ability to transparently sub-rate multiplex many lower-speed signals onto a single pipe would greatly increase efficiency, reduce fiber usage, and make transparent optical networks even more cost-effective.
To achieve multiprotocol subwavelength multiplexing, a framing mechanism is required that maintains the transparency of the optical signals and efficiently multiplexes several independent and different protocol streams.
The generic framing procedure (GFP) protocol approved by the International Telecommunication Union ITU-T provides a framing format capable of supporting a diverse array of higher-layer protocols over any transport network. GFP supports two protocol mapping modes: frame mode and transparent mode. These two mapping modes provide a means to map all existing and emerging protocols.
The frame mode is capable of adapting packet/PDU-oriented protocols. The current GFP standard defines a frame adaptation mechanism for IP/point-to-point protocol, any Ethernet protocol (10/100 Mbits/sec and 1/10 Gbits/sec), and resilient packet ring (RPR).
The transparent mode is defined for adapting time-sensitive protocols that require very low transmission latency. These protocols include any constant bitstream data transmission such as SAN protocols (Escon, Ficon, or Fibre Channel), video traffic, and even GbE. The current GFP standard defines transparent adaptation mechanisms for 8B/10B-coded data streams for Escon, Ficon, FC, and GbE.
Since the GFP protocol is not specific to any particular physical transport layer, it can be implemented over dark fiber. However, the most common implementation today is GFP over a SONET/SDH physical layer. GFP maps the protocols transparently into STS-1/STM-1 payload containers. With SONET/SDH, multiple timeslots can be bonded using the virtual concatenation (VC) method to form larger data containers, which effectively increases the transport data rates and container sizes available to the GFP protocol (see Table). The combination of GFP and VC provides a mechanism to transparently multiplex diverse protocols and data rates over a single optical signal.
A hybrid metro network using WDM, GFP, and SONET provides a network with the greatest flexibility and efficiency—at the lowest cost. The hybrid-network approach brings a WDM infrastructure to every site for delivering high-speed wavelengths and implementing dedicated "special" networks and GFP and SONET/SDH service interfaces for the greatest density of lower-speed services.
The networking devices used to implement a hybrid architecture will vary depending on where a site is located within a metro network. The metro network can be divided into two parts: access and core. There are unique requirements and characteristics in terms of interfaces, protocols, data rates, and traffic pattern for each part.
The access networks are collector networks to a hub site. There is a diversity of client protocols and data rates. The client signals can be TDM (T1, DS-3, OC-3), SAN (Escon, Ficon, FC), IP (Ethernet, Fast Ethernet, GbE), ATM, video, and many more. There are also some very-high-speed signals such as OC-48, OC-192, and 10-GbE.
It is highly desirable to transport each signal transparently on its own wavelength, but the high volume of signals would not make this approach cost-effective. The use of GFP would significantly reduce the number of wavelengths needed and lower the cost of delivering low-speed services. But the advantages of GFP are diminished as the data rate and the number of STS-1 timeslots needed per client signal increases. To minimize these costs, GFP should be implemented only on lower-speed services and where there are many diverse lower-speed signals to multiplex. That would be at the access network close to the client signals and at the edge of the metro core, where the expensive GFP interfaces and SONET/SDH multiplexing components can be used most effectively. Only the higher-speed signals should be given dedicated wavelengths and multiplexed where the added cost of GFP and SONET/SDH provides little benefit. Edge network devices that support transparent GFP multiplexing and WDM multiplexing would adapt low-speed service into GFP over SONET/SDH timeslots and wavelength multiplex the high-speed services over the same fiber infrastructure (see Figure 1).
The main function of the metro core network is to switch traffic to other parts of the network. Traffic from the access networks is switched to other access networks, to another metro core site, or to the regional/long-haul network. The typical optical signals received at a metro site are very-high-speed. These signals would be either densely packed GFP/SONET wavelengths with many low-speed client signals or very-high-speed transparent/dedicated wavelengths received from the access network or directly from a client interface.
These signals can be further aggregated, groomed, and switched for even greater wavelength, fiber, and timeslot utilization at each core transfer site in the network. The additional switching and grooming can be performed at both the STS-1/STM-1 level for GFP signals and wavelength level for wavelength services. A balance in the amount of SONET/SDH switching and wavelength switching is needed to prevent overbuilding the metro core capabilities and reduce costs. The cost of STS-1 switching on multiple OC-48 and OC-192 SONET/SDH signals is very expensive compared to the cost of transparently switching a wavelength.
In the hybrid metro network, SONET/SDH STS-1/STM-1 switching capabilities would be concentrated in the access network and at the edge of the metro core where the amount of grooming/switching needed is smaller and more easily forecasted. Meanwhile, the metro core sites would use more lower-cost wavelength switching to direct the signals to their destination sites and would be connected using mesh links to minimize the number of hops traveled (see Figure 2).
A hybrid network that combines transparent subwavelength multiplexing via GFP over SONET/SDH with transparent wavelength multiplexing via WDM provides the optimal use of bandwidth in the metro network. GFP and WDM eliminate the need to deploy "special" networks to accommodate the needs of a particular protocol while providing a common network infrastructure that is flexible and scalable.
Coleman Hum is a product manager at Meriton Networks (Ottawa, Ontario). He can be reached at firstname.lastname@example.org.