MAN Development and Technical Solutions

1 Problems with MAN and  Basic Requirements of  

   Next-Generation MAN
The primary problem that Metropolitan Area Network(MAN) is facing is the bandwidth bottleneck. On the user side, the emergence and development of low-cost Gigabit Ethernet has greatly improved the data rate of Local Area Network (LAN); on the long-distance network side, Wavelength Division Multiplexing (WDM) technologies expand the transmission capacity by orders of magnitude.   Therefore, MAN has become the bandwidth bottleneck of the whole network.
  Another problem with metro networks is the overlap of networks.
  Firstly, most telecom carriers deliver voice and private line services over SDH and circuit switches, and provide data services through SDH and separated frame relay, ATM and IP networks. This network separation requires separated network management and human resources, different network configurations and billing systems, and even different terminals. Due to the limitation of inertial thinking, organizational structure of telecom carriers and initial costs for network upgrade, this network-overlap situation will keep unchanged. However, from a long-term and overall point of view, it will inevitably increase the construction and operation costs of network expansion, and be more expensive to provide new services.
  Secondly, it is complicated and expensive for users to use services via different access technologies and routes.
  Thirdly, enterprise users require more customized applications instead of simple broadband connection, which requires the network to support complicated Layer-2 and Layer-3 functions. So a single service mode may reduce operation revenues and cannot retain subscribers.
  Lastly, the bottom layer of MAN mainly adopts the traditional SDH as a transmission platform. But SDH was designed for voice services with fixed bandwidth. While sending burst traffic, data transmission on SDH means low efficiency. Moreover, the change of bandwidth always means changing physical interfaces and even service patterns, which forces enterprise users to redesign and reconstruct their networks when they want to change services.
  At present, MAN has become a network capacity bottleneck and obstructs the further development of telecom networks. The next generation MAN has to effectively handle hybrid Layer 1/2/3 services to solve problems with MAN. Among those hybrid services, the proportion of services from different layers varies as time goes on. Therefore, the basic transmission networks should not only effectively support current
Layer-1 services with scalable capacity for service expansion, but also be able to provide Layer-2 and Layer-3 services, and to ensure a smooth transition to networks that support
Layer-3 services.

2 Solution 1: SDH Multiservice Platform
An SDH multiservice platform supports the transport of data services over SDH facilities, terminates multiple data protocols, and implements data transparent transmission or Layer-2 switching and local aggregation. Wherein, the point-to-point data transparent transmission means to transmit encapsulated data within a virtual container in an easy way and at a low cost. Besides, it guarantees users´ bandwidth and security isolation well, fit for leased line data services that require high QoS. However, this transmission mode has low bandwidth utilization and occupies a lot of network hardware resources. It does not support port aggregation, lacks flexibility, and requires manual configuration for each physical channel.
  The Layer-2 switching and local aggregation mode means that users access a network in a multipoint-to-point aggregation mode. Data packages are exchanged between Ethernet ports at the user side and virtual containers at the network side according to Media Access Control (MAC) addresses. This mode allows bandwidth sharing and port aggregation. User isolation and rate limitation can be achieved via the Virtual Local Area Network (VLAN). It makes use of the SDH ring and Rapid Spanning Tree Protocol (RSTP) to realize Layer-2 protection and bandwidth sharing on the ring, saving network resources and ports. However, since Layer-2 switching has the feature of competing for bandwidth, it is hard to ensure actual bandwidth and security for users. A  SDH multiservice platform with Lay-2 switching and local aggregation functions can greatly reduce service ports of nodes and network costs, and lighten the burden of Lay-3 switches and routers. It supports flexible networking. It is suitable for applications at the convergence layer and access layer. And it is also suitable for services that have low requirements for network security, such as online browsing and Video on Demand (VoD). In addition, with the integrated router function, it can further support data processing at Layer 3 to provide richer and more flexible data access functions.

  Such a Multiservice Transport Platform (MSTP) can save lots of individual service nodes and transmission nodes, simplify node structure, cut down equipment costs, simplify circuit allocation, and reduce the number of racks, spaces for equipment, power consumption and number of interconnections between racks. It can also shorten service provision time, improve network scalability, and save operation and training costs. In addition, it supports various data services. When integrated with Ethernet, frame relay, ATM and IP routing, it can improve the bandwidth utilization of Time Division Multiplexing (TDM) channels and reduce the number of equipment ports at central offices through statistical multiplexing and excess service order. This can optimize the existing SDH infrastructure. In addition, MSTP can provide all ports with any combination of Layer 1/2/3 services, no matter what types of physical interfaces are used.

  With increasing data services, MSTP is evolving from the current fixed-encapsulation and transparent transmission mode, which  supports simple data services, to a new-generation system, which can flexibly and effectively support data services. The latest achievements allow SDH to support standards of Generic Framing Procedures (GFP), Link Capacity Adjustment Scheme (LCAS), Resilient Packet Ring (RPR), and Automatic Switching Optical Network (ASON).

  The mapping of GFP is easy and flexible, and requires low overhead and has high efficiency. GFP helps the interworking of equipment from different manufacturers, supports various network topologies, makes statistical multiplexing for subscriber data, and has the OoS guarantee mechanism. In addition, it simplifies the data link mapping and de-mapping processes, which further reduces the complexity, size and cost of a receiver, thus making GFP especially suitable for high-speed transmission link applications, such as point-to-point SDH links, wavelength channels of Optical Transport Network (OTN) and dark optical fiber applications.

  The most strong point of LCAS is that the effective net load can be automatically mapped to an available Virtual Channel (VC). This means the bandwidth adjustment goes continuously. Therefore, it can speed up the bandwidth allocation, do no harm to services, and dynamically and automatically adjust system bandwidth with a soft protection mode and error tolerance mechanism when a system failure occurs.

  If the next generation MSTP can be integrated with the functions of VC concatenation, GFP, LCAS and RPR, as well as the automatic routing and allocation functions of intelligent optical core networks, it will be able to support data services more flexibly and effectively. Furthermore, it will extend the intelligence of intelligent optical core networks to network edges, and enhance the intelligence of the whole network.

   
  In general, MSTP serves best as a convergence node at a network edge, supporting hybrid traffic, especially that with dominant TDM traffic. New carriers who lack telecom infrastructure can use MSTP for interoffice communication, and even internal communication of large-scale enterprises. As for those carries with a large-scale SDH networks, MSTP can support packet data services, enhance service delivering capability, reduce operation costs, and help realize the transition from circuit switching networks to packet switching networks more flexibly and effectively.

  However, this solution has some shortcomings. Firstly, it requires network synchronization and strict limits on jitter, thus bringing on more equipment costs. Secondly, it is hard to create services very flexibly. Thirdly, the bandwidth utilization is inefficient when fixed time slots are used to support data services, and the functions of available data services are not rich enough. Lastly, it costs much to manage connection-oriented and connectionless networks simultaneously. From a long-term point of view, once data services occupy the leading position in the future, the solution won´t be the best one, and may be replaced by more effective solutions.

3 Solution 2: Ethernet  Multiservice Platform
The Ethernet technology originates from LAN. It is simple and well known after years of application. With the technology, the time for service assignment can be reduced to a few hours or days. It is a standard technology with excellent interoperability. It is inexpensive and widely supported by hardware and software. As a bearing technology, Ethernet is independent of any media, so it can be connected with various transmission media such as copper twist-pairs, cables and optical fibers, which saves rewiring costs.

  From the respect of network structure, Ethernet is an end-to-end solution. Different parts of a network work as a whole to handle Layer-2 switching, traffic and service allocation, which simplifies the network structure and avoids format conversion at network edges. With good scalability, the Ethernet capacity can be set at three levels, i.e. 10 Mb/s, 100 Mb/s, and 1 000 Mb/s. At a network edge, the bandwidth granularity can be increased from 64 kb/s to 1 Mb/s, and up to 1 Gb/s by changing traffic policing parameters. Nowadays, a 10 Gb/s Ethernet has been available.

  Since the same Ethernet system can be deployed on different layers of a network, network management is significantly simplified. In addition, many users have been familiar with Ethernet, which helps develop new services. Therefore, with the Ethernet MAN, service providers can quickly and economically offer high-speed data transmission and application services on users´ demands.

  Generally, the Ethernet Multiservice Platform is best fit for applications with dominant IP data traffic. It can be used as an independent IP MAN in middle-size and small cities with great demands for IP services. It can also be used as the convergence and access layer of an IP MAN in large or middle-size cities with great IP data traffic.  The high-end router is the core of Ethernet Multiservice platforms. As IP  traffic keeps growing, the platform will have more and more applications in MAN.

   However, it does not mean that there are no problems. Firstly, although QoS is not a problem when Ethernet is used for LAN, when Ethernet is used for public telecom networks, its QoS is required to meet the requirements of different services for different QoS and Service Level Agreement (SLA) mechanism. At present, Ethernet can neither guarantee point-to-point jitter and delay, nor provide standard QoS guarantee for real-time services. It also fails to provide the billing statistics function that is necessary for multiple users´ sharing nodes and networks.

  Secondly, Ethernet was originally designed for enterprise LANs, so it lacks of security assurance mechanisms. When it is deployed for MAN and Wide Area Network (WAN) that deliver services on the same infrastructure to a large number of subscribers, a new security mechanism is required.

  Thirdly, Ethernet was originally used for small LANs and weak in network Operation, Administration, Maintenance and Provisioning (OAM&P). However, when used in public telecom networks, Ethernet is required to be powerful in OAM&P and have network-level management capability and profitable business modes.

  Fourthly, the optical interfaces of an Ethernet exchange are connected in a point-to-point mode, so it saves the transmission equipment. As a result, the lack of built-in fault location and performance monitoring makes it hard and expensive to diagnose and recover network faults, especially faults in large complicated networks. Ethernet has no built-in self-protection functionality, and is protected mainly by routers. It costs Ethernet at least 1s to redirect data streams, so it is impossible to transport carrier-class voice and data streams at current status.
Fifthly, with the expansion of networks and increase of nodes, the cost of  fiber cable rockets. Whether it is economical or not for complicated large carrier-class networks to use Ethernet still remains unknown.

  Lastly, although Ethernet is a mature technology for LAN, it still needs time to validate its feasibility when it serves large carrier-class public networks. In a word, only after the above problems are solved, can Ethernet be deployed as a multiservice platform in a large public network environment, delivering various carrier-class services.

4 Solution 3: RPR Multiservice Platform
In order to extend Ethernet into a carrier-level core network, some inherent problems of Ethernet should be solved, to which RPR is one of the solutions. As a middle-layer enhanced technology, RPR adopts a packet-switching mechanism based on Ethernet or SDH. In an access-sharing mode, RPR directly sends IP packets to data frames of Layer 1 or bare optical fibers via a new MAC layer, dispensing with the need for packet disassembly and reassembly. This boosts the network switching capability, and improves network performance and flexibility. RPR can work on both SDH and Gigabit Ethernet, and on bare optical fibers as the interface board of routers as well. Stand-alone RPR equipment was previously deployed on Ethernet. But now there is a trend to place it on SDH as an embedded function entity of the next generation MSTP to make full use of their advantages.

  Unlike the Ethernet, which handles packet disassembly and reassembly, sequencing, grooming and processing at each node, RPR simplifies data packet processing by forwarding transient IP packets directly, obviously enhancing the switch processing capability. It is suitable for packet services, and can also ensure QoS of circuit switched services and private services (with 50 ms protection switching).

  With the automatic topology recognition ability, RPR can recognize any topology transformations at Layer 2, improving its self-healing capability. It supports plug-and-play functionality, avoiding errors brought by time-consuming manual configuration. With an effective two-fiber dual-direction ring topology, it can dynamically multiplex various services in dual directions of the ring, and reserve bandwidth and QoS for each user or service. So it can maximize the bandwidth utilization of optical fibers, simplify network configuration and operation, and speed up service deployment. Besides, RPR has a good bandwidth balance and congestion control mechanism.

  A predominant characteristic of RPR is that it adopts an embedded control layer to provide a number of new functions. RPR costs more than SDH and less than Gigabit Ethernet. The more data interfaces there are, the closer is its cost to that of Gigabit Ethernet, whereas the cost is close to that of SDH. Generally, this technology is best suitable for applications with dominate data traffic and TDM traffic.

  Since RPR has outstanding convergence functionality and optimized data access capability, it can be applied to the access layer of MAN, especially when Ethernet services demand most bandwidth.
However, a new MAC layer increases the system cost of RPR. Besides, not supported by any cross-ring standards, RPR information cannot be transported between rings. So it is weak in independently setting up a large network, and cannot realize complicated network topologies such as the tangent ring, intersection ring and chain of rings and provide end-to-end services. Due to access sharing, the scalability of RPR is limited. However, combined with Multiprotocol Label Switching (MPLS), it can implement cross-ring services by transporting cross-ring traffic on a common MPLS channel.

5 Solution 4: WDM Multiservice Platform
Along with the development of technologies and services, the WDM technology used in the long-distance transmission field is expected to be applied for MAN. Such application needs technical modification according to MAN´s special environment.

  What benefits can this solution bring? Firstly, WDM helps enlarge the network capacity by ten even and hundred times. Besides, it can provide WDM ring protection.
 

  Secondly, WDM allows network carriers to deliver transparent services based on wavelength. Users can forward signals in any formats, instead of those in SDH format only. On one hand, a WDM system at the edge of MAN should be able to flexibly and quickly provide users with interfaces of various data services of different rates and signal formats. This requires its optical interfaces automatically receive and adapt to all signals ranging from 10 Mb/s to 2.5 Gb/s, including signals from SDH, ATM, IP, Gigabit Ethernet and optical Channels. One the other hand, a WDM system that is used as a core system of MAN is expected to support SDH signals ranging from 10 Gb/s to 40 Gb/s and Ethernet signals in the future.

  Lastly, a WDM system has scalable wavelengths. New wavelength can be freely added without any influence on the working wavelength. So the system can rapidly deliver new services by simply adding new wavelengths, greatly improving network scalability and carriers´ market competitiveness.

  However, WDM Multiservice Platform costs much now, especially when the transmission distance is long and optical amplifiers are required. It is necessary to develop low-cost optical amplifiers. Since there are few users and applications requiring for the whole wavelength bandwidth, the platform is mainly used at the core layer of MAN. It is especially fit for long-distance applications with great demands for capacity expansion. Therefore in order to extend WDM´s application to the edge of a network, it is necessary to further develop sub-rate multiplexing technologies that allow different traffic and protocols to share a common wavelength and improve capacity utilization.

  The Coarse Wavelength Division Multiplexing (CWDM) technology has been developed to lower the cost of WDM multiservice platform. It has 3 typical combinations of wavelength, i.e. 4, 8 and 16 wavelengths, with respective coverage ranges of 1 510-1 570 nm, 1 470-1 610 nm and 1 310-1 610 nm. The interval of adjacent two wavelengths is 20 nm. The filter pass-band is about 13 nm, allowing wavelength drift of 6.5 nm. Thus requirements on the laser module are greatly relaxed. The laser module of a traditional DWDM system requires the wavelength precision less than 0.1 nm, but the wavelength precision of a CWDM laser can reach 2-3 nm. Even a DVD production line can be used to produce the laser module of a CWDM system, so the cost can be greatly reduced. In addition, a CWDM system has such a low requirement on the wavelength precision of its laser that coolers and wavelength locking devices are unnecessary. Therefore, the CWDM laser module can be smaller and consumes less power. It can be encapsulated in a simple coaxial structure, costing only 1/3 of the cost for traditional dishing encapsulation. As for the filter, a typical thin 100 GHz film filter needs 150 plating films, while a CWDM filter with 20 nm spacing needs 50 only. The production ratio is accordingly improved, and the cost is expected to lower by half, at least.

  In short, a CWDM laser requires much less on output power, temperature sensitivity, dispersion tolerance and encapsulation than a DWDM laser. The requirements on filter have also been lowered. Therefore, it is hopeful to greatly cut its cost down. A CWDM system with 8 wavelengths will be used first because its spectrum arrangement avoids the absorbing peak around 1 385 nm, which makes any optical fibers applicable.

  As for service applications, the CWDM transceiver has been used as Gigabit interface converters and small plug-and-play devices. It can be inserted to Gigabit Ethernet switches and optical fiber channel switches, allowing users to select a wavelength by themselves. Its dimension, power consumption, and cost are significantly less than those of a DWDM transceiver. At present, a 100 GHz Gigabit interface converter is available, and a 50 GHz one is expected to appear soon. A user can first deploy a CWDM system to meet demands for services, and just replace the CWDM transceiver with a DWDM one without changing other parts when traffic requires more wavelengths. This simple replacement will bring a smooth upgrade to a system with tens to hundreds of wavelength channels.
Generally speaking, a WDM multiservice platform is a general and long-life solution for a MAN lacking optical fiber resources or for the core layer and even the future convergence and access layers of a large MAN. It is best fit for the convergence and access layers of a MAN.

6 Conclusions
The four solutions, aiming at different MAN application environments, work together to build up a complete MAN solution. For most carriers, it is a stable and sustainable strategy to select a SDH multiservice platform at present, taking into consideration of both existing SDH facilities and demands for data services.

  An RPR multiservice  platform has advantages when applied at the edge of a network. An Ethernet  multiservice platform will be a major solution when IP services are dominant. Once traffic is heavy enough, a  WDM multiservice platform will play a leading role at the core and convergence layers of MAN.

  As a whole, SDH multiservice will play a center role in the near future.
It is believed that, with rapidly increasing IP services and commercially deployed 3G services in sight, a dynamic and flexible MAN with broad bandwidth will be a necessity to the network development. It will also bring an important market opportunity.

Manuscript received: 2003-10-10