STRUYVE LAYOUT

4 downloads 0 Views 67KB Size Report
planning, network and service management, telecom soft- ware, internetworking, network protocols, QoS, and so on. He has published over 300 papers in the ...
WDM OPTICAL NETWORKS: A REALITY CHECK

Application, Design, and Evolution of WDM in GTS’s Pan-European Transport Network Kris Struyve, Nico Wauters, and Pedro Falcao, GTS Network Services Peter Arijs, Didier Colle, Piet Demeester, and Paul Lagasse, Ghent University — IMEC

ABSTRACT A pan-European transport network must cope with enormous traffic inflation by introducing appropriate network solutions. WDM technology is the solution adopted by GTS, driven by capacity requirements, infrastructure availability, flexibility, and cost. The deployment of this new technology presents specific technical issues, but more important, it lays the foundation of future optical networking. In this article we describe the motives underlying the deployment of WDM technology in the long-haul and short-haul areas of GTS’s operational transport network. Next, after presenting the use of WDM as the multiservice platform for SDH and IP overlay networks as well as native WDM services, we focus on the design of SDHover-WDM and IP-over-WDM networks. Finally, we discuss future WDM networking and management requirements to better serve the needs of a pan-European carrier such as GTS.

INTRODUCTION Telecommunications traffic continues to grow at a very fast pace. This is accelerated through the increasing volume of data and mobile traffic and, especially in Europe, through the recent liberalization of the telecommunications market, where competition is having a positive impact on the price of network capacity. This trend is especially experienced by a carrier such as Global TeleSystems (GTS), whose pan-European transport network, formerly known as Hermes Europe Railtel (HER), is situated in the network core of its carrier customers. The solution adopted by GTS to meet the ever-increasing traffic requirements is based on a combination of WDM, SDH, and IP transport technologies. First, GTS uses wavelength-division multiplexing (WDM) to multiplex several wavelength channels on a single strand of fiber, thus overcoming fiber congestion. Second, synchronous digital hierarchy (SDH) technology offers the capacity granularity GTS’s carrier customers demand today and offers the possibility to protect these services against network outages. Third, GTS uses an IPover-WDM transport network to offer highcapacity Internet transit services to Internet service providers (ISPs) in Europe.

114

0163-6804/00/$10.00 © 2000 IEEE

WDM MEETS THE CAPACITY CRUNCH WDM IN THE LONG HAUL In 1995, long-haul carriers in the United States started deploying point-to-point WDM transmission systems to upgrade the capacity of their networks while leveraging their existing fiber infrastructures [1]. Since then, WDM has also taken the long-haul market by storm in Europe. There are several drivers for a pan-European carrier like GTS to introduce WDM in its longhaul transport network. WDM technology allows GTS to cope with ever-increasing capacity requirements while postponing the exhaustion of fiber and increasing the flexibility for capacity upgrade [2]. The most prevailing driver, however, is the cost advantage of the WDM solution compared to competing solutions such as spacedivision multiplexing (SDM) or enhanced timedivision multiplexing (TDM) to upgrade the network capacity. To better understand this cost difference, we will first highlight the key technological differences. The “open” WDM solution, illustrated in Fig. 1, makes use of transponders in WDM terminal multiplexers (TMs) and inline optical amplifiers that are shared by multiple wavelength channels. The transponder is in essence a 3R opto-electro-optic (O/E/O) converter that converts a G.957 standard-compliant optical signal into an appropriate wavelength channel (and vice versa) while repowering, reshaping, and retiming the signal electrically. The SDM solution uses multiple fiber pairs in parallel, each equipped with SDH regenerators instead of multiple wavelengths sharing the same inline optical amplifier. Upgrading to higher TDM rates (e.g., from 2.5 Gb/s STM-16 to 10 Gb/s STM-64) is only a short-lived solution since transmission impairments such as dispersion do not scale well with increasing TDM rates, especially on standard single-mode fiber. A case study by GTS has demonstrated that long-haul point-to-point WDM systems are clearly a more cost-effective solution than SDM, even for as low as three channels of STM-16. Figure 2 illustrates two link cost comparisons for the initial core of the GTS transport network consisting of 5000 fiber km in seven countries with an aver-

IEEE Communications Magazine • March 2000

age distance of 300 km between two access cities. Note that the 100 percent cost reference point in Fig. 2 corresponds to the cost of deploying one STM-16 channel, including fiber cost. Two conclusions can be derived from Fig. 2: • Figure 2a shows that if only transmission and regeneration equipment costs are considered (i.e., SDH regenerators in the SDM case, and WDM TMs with transponders and inline optical amplifiers in the WDM case), the initial link cost of using WDM technology is more than double that of SDH. However, the WDM solution is more cost-effective for the deployment of three channels and more in the network, because of the shared use of the inline optical amplifier. • Figure 2b shows that if, in addition to Fig. 2a, fiber cost is also considered, the cost advantage of the WDM case becomes even more evident and is amplified as the number of channels increases. The WDM solution is more cost-effective for the deployment of three channels and more in the network.

1 2

1

Inline optical amplifier W D M

W D M

T M

T M

40

2

G.957 standardized SDH interface 40

Inline optical amplifier Transpoder

■ Figure 1. An open point-to-point WDM system. Equipment using G.957 interfaces interconnect to the WDM terminal multiplexers (TMs) via transponders. GTS point-to-point WDM systems can transport up to 40 wavelength channels. and protected using Drop&Continue. These increased intracity capacity requirements have led to the deployment of WDM in the short-haul section of GTS’s pan-European transport network. The main reason WDM is preferred over SDM is because fibers in a city have to be leased from a third party or built. Leasing or building city fiber is not only an expensive process, it is also a less flexible approach to upgrade capacity. In a dynamic environment, where traffic distributions and volumes evolve rapidly, the amount of fiber to be leased or built is hard to predict in advance. Therefore, using WDM technology has clear flexibility advantages because wavelength channels can be activated in a very short time. Although specific short-haul WDM systems are available, GTS preferred to use the same type of WDM system as for its long-haul network. While short-haul WDM systems are less expensive than their long-haul counterparts, because low-cost optical components can be used, they lead to a heterogeneous network, which is not preferred for several reasons. First, using two different systems leads to an increased operational and management cost. For instance, a heterogeneous network requires more spare equipment parts than a homogeneous network. Second, the interworking between two different systems might pose problems. For instance, bottlenecks can occur because short-haul WDM sys-

WDM IN THE SHORT HAUL In the previous section we discussed the benefits of WDM in long-haul networks. Because of the limited distances in short-haul networks, regenerators are not necessary and optical impairments have less impact, so the benefits of WDM are less clear than those of SDM or enhanced TDM solutions. However, fiber exhaustion and low-cost optical components are now also driving WDM in the metropolitan area. The initial short-haul application targeted by GTS is related to the interconnection of multiple points of presence (POPs) within the same city. As shown in Fig. 3, GTS has at least two POPs per city where customers can interconnect. With dual node interconnection techniques such as Drop&Continue [3], customer networks can be interconnected with the GTS transport network via two different POPs. This results in a very secure architecture that can even survive POP failures without any traffic impact. Thus, the traffic flow between two POPs in a city consists not only of traffic that passes through the city, but also of traffic that is terminated in the city

150%

500% SDH regenerators (equipment only) WDM (equipment only)

SDH regenerators + N x fiber WDM +2 fiber

400%

100% 300% 200%

50%

100% 0%

0% 1

2

3

4 5 6 2.5 Gb/s channels

7

8

(a)

1

2

3

4 5 6 2.5 Gb/s channels

7

8

(b)

■ Figure 2. WDM versus SDM cost analysis for average long distance links.

IEEE Communications Magazine • March 2000

115

Long-reach WDM link

PoP1 Short-reach WDM link PoP2

To Amsterdam

To Brussels

at the optical layer solely, and thus enables the deployment of an SDH client layer with a topology different than the fiber layer. Intermediate transponders are still required to refresh the distorted optical signal and allow signal monitoring at regular points. Figure 4 illustrates the concept of so-called SDH overlay rings in which ADMs are only installed at locations where the STM-16 streams need to be accessed. In the future, this wavelength pass-through can also be achieved with more advanced optical network elements.

IP OVER WDM

To trans-Atlantic cable

To Paris

■ Figure 3. Part of the GTS transport network in the London area where multiple POPs are interconnected via WDM systems.

tems typically support fewer wavelengths than long-haul ones.

WDM AS A MULTISERVICE PLATFORM The introduction of point-to-point WDM systems not only boosts network capacity. It also provides the foundation of an optical network supporting multiple “client overlay networks” and transporting novel optical services. This section briefly describes GTS’s use of WDM as a multiservice platform for SDH and IP overlay networks as well as native “lambda services.” See [2] for a more in-depth discussion.

SDH-OVER-WDM Whereas the introduction of point-to-point WDM systems is a viable and cost-effective solution as of today, legacy transport technologies such as SDH are not expected to disappear overnight. Present point- to-point SDH services at E1/VC-12, E3/T3/VC-3, or E4/VC-4 rates continue to contribute a substantial amount of network traffic. In the near future, SDH remains a key technology for provisioning, managing, and protecting these services. Indeed, the flexibility SDH offers related to circuit configuration and its protection capabilities have currently not reached a sustainable level of maturity in the optical layer. In addition, SDH cross-connecting allows improvement of optical channel utilization by consolidating SDH traffic streams stemming from multiple origins. Initially, the SDH network can be supported by simple point-to-point WDM channels between adjacent nodes. However, because the amount of traffic passing through the nodes is ever increasing, definite opportunities exist to process part of the traffic optically, alleviating the more expensive electronic processing, that is, reducing the number of SDH add-drop multiplexers (ADMs) and digital cross-connects (DXCs). The “pass-through wavelength functionality” achieved by back-to-back interconnection of transponders at the WDM TMs allows bypassing certain nodes

116

Since mid-1999 GTS has been operating a panEuropean IP-over-WDM core network to cope with the huge increase of Internet and IP transit services. Prior to that IP routers were interconnected to SDH ADMs via STM-1 tributary cards. This approach demonstrated unscalability because of the difficulties in load-balancing IP traffic between two IP routers among several separate STM-1 circuits. In addition, the existing SDH network of GTS does not support concatenated payloads (e.g., VC-4-4c, VC-4-16c) as required by IP routers for connectivity beyond 155 Mb/s. For these reasons, since mid-1999 IP gigabit switch routers have been directly interconnected to WDM TMs via optical STM-16 aggregates, thus avoiding any SDH equipment in the core network and reducing the unit cost. Similar to SDH over WDM, the pass-through wavelength functionality can be used to optically bypass certain nodes and optimize the logical IP layer topology in the function of IP traffic distribution.

LAMBDA SERVICE The GTS “lambda service” is an optical transport service targeted at customers who have very high capacity requirements (e.g., 622 Mb/s or 2.5 Gb/s), use SDH interfaces with concatenated payloads (e.g., VC-4-4c or VC-4-16c ), or want full manageability of their own SDH, IP, or asynchronous transfer mode (ATM) network — that is, transparency of SDH data communication channel (DCC) bytes. The customer connects directly to GTS’s optical network by means of a transponder on the WDM TMs, bypassing all SDH equipment. Back-to-back interconnection of transponders enables GTS to provision the lambda service end to end to the customer. The transparency of the lambda service is determined by the transponders since they typically support only a discrete set of digital client signals (e.g., STM-4 and STM-16).

OVERLAY NETWORK DESIGN This section focuses first on the design of SDHover-WDM networks. Two key design issues relate to the appropriate choice of survivability strategy and the dimensioning of the network. Next, some design concepts for IP-over-WDM networks are provided based on GTS’s experience and future plans.

SDH-OVER-WDM NETWORK DESIGN Choice of Survivability Strategy — An important requirement for GTS is the survivability of its SDH-over-WDM network in order to cope with network failures. Because recovery

IEEE Communications Magazine • March 2000

mechanisms in the WDM layer are still in a premature stage, all recovery actions are currently executed at the SDH layer. Three broad categories can be considered for survivable SDH architectures: end-to-end path protection, interconnected protection rings, and restoration. Restoration was not considered because of the lack of relevant standards, slower restoration times, increased need for expensive cross-connecting devices, and unpredictable network configuration effects. End-to-end path protection, based on either subnetwork connection protection (SNC-P) or trail protection [3], uses two diversely routed paths between both endpoints of a connection. In case of a failure on the main path, traffic is switched to the protection path within 50 ms. Since the failure probability of a path is typically proportional to its length, and some paths may span several thousands of kilometers in GTS’s pan-European transport network, the possibility of double failures affecting both the main and protection paths cannot be neglected. For this reason end-to-end path protection was not considered a viable alternative for GTS. On the other hand, connections routed across multiple interconnected protection rings have much higher availability, because protection independence is achieved. Indeed, failures occurring in one ring are restored within that ring only without affecting the other rings, such that certain multiple failures that occur concurrently in different rings can be recovered from. Two kinds of SDH rings can be considered: SNC-P rings and MS-SPRings. SNC-P rings work along the same principle as end-to-end SNC-P path protection, but within a ring structure. Alternatively, MS-SPRings are rings that dedicate 50 percent of their capacity for protection purposes; in case of a failure, all affected traffic is looped back along the other side of the ring using this protection capacity. Traffic, which is routed across multiple rings, can be protected against a failure of the ring interconnection gateway by using Drop&Continue [3], which requires two gateway nodes between both rings. MS-SPRings have been proven to use capacity more efficiently than SNC-P rings for most traffic patterns [3], because the protection capacity can be shared among multiple connections. However, in an interconnected ring network, the cost benefit of MS-SPRings is not so strong. A case study comparing the use of SNC-P rings versus MS-SPRings in the GTS transport network revealed that the relative cost savings using MS-SPRings were about 10 percent. This limited cost benefit of MS-SPRings can be explained by the fact that the traffic pattern on the rings is not in favor of MS-SPRings because a limited number of nodes on each ring attract most of the traffic, such as large business centers and gateway nodes between rings. The influence of the latter is even augmented when Drop&Continue is used. In addition, SNC-P rings are simple to operate, flexible regarding unpredicted demand dispersions, and have slightly faster restoration times compared to MS-SPRings. Therefore, SNC-P rings were selected for the initial rollout of GTS’s transport network, but this does not preclude future deployment of MSSPRings.

IEEE Communications Magazine • March 2000

Inline optical amplifier (a) Fiber topology Node

(b) WDM topology

Pass-through of wavelength through back-to-back interconnection of transponders

WDM TM

SDH ADM (c) SDH topology SDH overlay ring

Node without SDH equipment for this overlay ring Tributary connection

■ Figure 4. Nodes in an SDH overlay ring. Note that a, b, and c give a different view of the same network: a) focuses on the physical infrastructure (fibers and optical amplifiers); b) on the WDM equipment and how it is interconnected; and finally, c) highlights the SDH equipment.

Network Dimensioning — Once the survivability strategy has been chosen, the next step is to dimension the network based on a traffic forecast for the next several years. Efficient network dimensioning is of paramount importance for making proper decisions on equipment procurement and installation. The process of dimensioning a network based on interconnected protection rings can be divided into two major phases: ring selection and routing, and optimization of overlaying rings. The outcome of the ring selection and routing phase is a suitable set of interconnected rings such that all nodes in the network are reachable at minimal cost. For this purpose a choice first has to be made between all possible ring positions within the fiber topology. This choice is limited due to the maximum ring length allowed for availability reasons. Next, different ring combinations have to be evaluated based on the traffic matrix. This cost evaluation uses a routing algorithm, which optimizes the amount of intra-ring traffic in order to efficiently use the ring capacity and limit the amount of inter-ring traffic. In the second phase, SDH overlay rings with the same fiber topology, so-called stacked rings, can be optimized by replacing ADMs with a pass-through wavelength in some nodes of these rings where no traffic needs to be terminated. Hence, by routing the connections in an opti-

117

Fiber topology

Logical topology

■ Figure 5. Three stacked SDH rings with different ADM configurations.

illustrates the target design of GTS’s pan-European IP-over-WDM core network. In order to cope with router failures or allow router upgrades without affecting network performance, the typical interconnection design consists of two core IP routers in tandem on the same ring that are optionally located in physically separate POPs. High-speed customers are directly connected to the core IP routers, while edge IP routers are deployed to aggregate lower-speed customers. The SDH interconnection between the customer or edge IP router on one side and the core IP router on the other side can be made resilient using a redundant parallel SDH interconnection providing resilience at either the IP or SDH layer. The first solution implies two working IP links in parallel and uses IP rerouting. The latter implies two SDH links in a working/protect configuration and uses linear 1+1 SDH multiplex section protection (MSP).

FUTURE NETWORKING AND MANAGEMENT REQUIREMENTS NETWORKING REQUIREMENTS

mized way on the stacked rings, the amount of ADMs required on all stacked rings can be minimized, thereby using the ADMs more efficiently and reducing investment costs, at the expense of reduced flexibility in case of traffic fluctuations. Figure 5 shows an example of three six-node SDH rings, each configured with only four ADMs. Note that it is still possible to route traffic between any two nodes on the three rings. An optimal solution to this problem is presented in [4] for MS-SPRings in which savings of 40 percent and more have been achieved in terms of the required amount of ADMs. Using a heuristic optimization method, we found out that for an interconnected ring network 33 percent of the ADMs could be eliminated and replaced by back-to-back transponders. Since transponders remain an expensive device, however, the total node cost decreased only by 19 percent. This makes the business case for lowcost reconfigurable optical ADMs very manifest, because in such a configuration the cost savings could be more explicit.

IP-OVER-WDM NETWORK DESIGN The initial rollout of GTS’s IP-over-WDM core network is based on a ring architecture with a number of core IP gigabit switch routers interconnected via a minimal set of STM-16 wavelength channels. The deployment of this core European ring is being followed by the introduction of peripheral rings collecting and distributing traffic from Southern, Eastern, and Northern Europe. Additionally, the core European ring will be upgraded to STM-64 to cope with traffic growth, and pass-through wavelength functionality will be used to increase network meshing. For example, it is foreseen to improve the connectivity from the peripheral nodes to the core gateway nodes that are linked to the U.S. Internet. Traffic requirement analysis and network dimensioning with excess capacity allows congestion avoidance in the IP-over-WDM core network even with outages in the server transmission layer. Figure 6

118

Currently, the main application for WDM systems is upgrading capacity on point-to-point fiber links. The next step is all-optical processing of traffic in a node in order to reduce the amount of processing resources in the client layers, thereby reducing both cost and complexity. Pass-through wavelengths, as explained previously, as well as optical ADMs (OADMs) and optical cross-connects (OXC) can achieve this. The latter two are more sophisticated and offer other benefits in the areas of provisioning and survivability. Equipment Evolution — Current OADMs are in fact inline optical amplifiers with interstage access, such that they can be extended with a module able to add/drop wavelengths via a similar type of transponder as used in WDM TMs. The other wavelengths, called transit wavelengths, pass the OADM transparently without any O/E/O conversion. Today’s OADMs, however, can only add/drop a fraction of the number of wavelengths supported by the WDM TMs (e.g., four or eight wavelengths out of 40). This reduces their application areas considerably to places where it is certain that the amount of transit traffic will be much larger than the amount of add/drop traffic. Since traffic distributions and volumes are very unpredictable, only a limited number of these OADMs are deployed in nodes with minimal risk of blocking the access. Another disadvantage of current OADMs, and of transparent optical processing in general, is their limited performance monitoring capabilities. Indeed, at present there are no simple optical performance monitoring techniques without regeneration available, which can be installed cheaply in large numbers, to assess the performance of the digital payload. OADMs terminate only two or four fiber pairs, whereas OXCs terminate tens of fiber pairs. OXCs can be all-optical or opaque. Alloptical OXCs cross-connect optical channels from an inlet fiber to an outlet fiber through an optical matrix without O/E/O conversion. Trans-

IEEE Communications Magazine • March 2000

parency is their major advantage. They can, for instance, cross-connect optical channels transporting STM-16 as well as STM-64 signals. However, scalable large-size nonblocking all-optical OXCs, in particular featuring wavelength conversion, are hard to develop and manufacture. Consequently, they are currently limited in port size (e.g., 16 wavelengths). Moreover, the above statement on limited performance monitoring capabilities also applies to all-optical OXCs. Due to transmission impairments, transparent optical processing without intermediate regeneration at regular places seems not viable in the near term for long-haul networks. In such a situation, it is recommended to regenerate the signals in places where some wavelengths have to be terminated anyway to give customer access, leading to a so-called opaque optical network [5]. A suitable place to regenerate transit signals is at opaque OXCs. These have either an optical cross-connect matrix with transponders at the boundaries or an entirely electrical cross-connect matrix. The main advantages of working opaque are regeneration, adequate performance monitoring capabilities, and isolation of optical impairments as opposed to cascading noise and crosstalk. Also, if the opaque OXC is interconnected with WDM TMs via standardized interfaces, it is in principle possible to tie different WDM point-to-point systems with different numbers of wavelengths and even from different suppliers to a single opaque OXC. The downside of working opaque is the limited transparency. For instance, opaque OXCs designed for STM16 wavelength payloads may not support upgrading to STM-64 wavelength payloads. Provisioning Requirements — Dynamically reconfigurable OADMs and OXCs will make it possible in the future to provision optical channels remotely. At least three ways of provisioning can be identified, corresponding to semi-permanent, soft-permanent, and switched optical channels. First, similar to SDH paths (e.g., VC-4, VC-12), the carrier instructs the WDM network management system (NMS) to establish a semi-permanent optical channel. The carrier maps out the route through the WDM network and configures the optical network elements appropriately. Second, similar to ATM soft-permanent virtual circuits (PVCs), the carrier configures only the ingress and egress interfaces of a soft-permanent optical channel while the WDM NMS automatically routes and configures it across the WDM network. Third, similar to ATM switched virtual circuits (SVCs) and voice circuits, a switched optical channel is established fully automatically using signaling protocols (e.g., using multiprotocol label switching, MPLS, techniques for IP-over-WDM services [6]). Support for semi- permanent as well as soft-permanent optical channels is mandatory. Support for switched optical channels requires further analysis, although it may drastically alter today’s static and cumbersome process of provisioning transport capacity by increasing flexibility and dynamics. Survivability Requirements — Besides provisioning, OADMs and OXCs will also enable

IEEE Communications Magazine • March 2000

Core European ring Gateway nodes to the U.S. Internet

US

Peripheral European rings IP over WDM (10 Gb/s) IP over WDM (2.5 Gb/s) IP over SDH (155 Mb/s)

■ Figure 6. Target design of GTS’s pan-European IP-over-WDM core network.

optical recovery to replace or complement protection in the electrical client layers. GTS envisages following two major applications of optical recovery [2]. First, as “first line of protection,” fast optical recovery is a must to protect the GTS lambda service as it cannot be protected at the SDH layer. For this application optical multiplex section (OMS) and optical channel (OCh) protection mechanisms seem inherently attractive [7]. Second, GTS envisages using optical recovery as a second line of protection for its SDH and IP overlay networks to make the network survivable during repair time of a single failure and also to recover from multiple failure occurrences. Since it is recommended to avoid unpredictable interaction between optical recovery on one hand and SDH and IP recovery on the other, GTS intends to implement them at different speeds. This could be achieved by fast first-line protection in the SDH layer activated automatically, and automatic first-line rerouting at the IP layer on one hand and slower second-line recovery in the optical layer activated by operator intervention on the other hand.

OPERATIONAL AND MANAGEMENT REQUIREMENTS Today’s NMS for the WDM network is functionally restricted to the element management layer (EML). Point-to-point WDM systems are managed as individual network elements (NEs) from a central network operations center (NOC). To cope with a catastrophic outage of the NOC, a redundant NMS is installed in a remote standby operations center. With the advent of new optical NEs, such as dynamically reconfigurable OADMs and OXCs, a more sophisticated NMS is required to operate and manage future WDM networks. Therefore, it is mandatory that the WDM NMS supports the network management layer (NML) as well. The NML manages the WDM network as a whole, while the EML manages the WDM NEs individually. Moreover, to offer WDM services such as the GTS lambda service to carrier customers, it is required to implement trouble tick-

119

WDM technology not only increases transport capacity, it also supports architectural concepts like SDH and IP overlay networks and emerging native optical services.

eting systems, service availability monitoring and reporting systems, as well as other operational support systems (OSS). These OSSs form the service management layer (SML) of the WDM NMS. Present operational WDM networks based on point-to-point systems reveal fault and performance management requirements that are of paramount importance for the future of optical networking. The first priority is support of fault detection, notification, and isolation processes. In the absence of such processes, and due to interworking effects such as “cascaded loss of signal (LOS) defects” [2], a network outage such as a fiber cut or a WDM equipment failure causes an avalanche of alarms, overwhelming the operators in the NOC. The second priority is the connectivity and quality supervision of OChs end to end across multiple point-to-point systems, OADMs, and OXCs in a pan-European WDM network. At present, to maintain signal quality and detect degradations in time, the NMS monitors a number of physical properties of the pointto-point WDM system such as laser current, laser power, laser temperature, and amplifier power. Equipment alarms are raised when the physical properties exceed preconfigured thresholds. Also, the point-to-point WDM system detects SDH LOS and loss of frame (LOF) defects, and monitors the SDH overhead nonintrusively. Typically, it nonintrusively monitors the B1 and J0 bytes in the SDH regenerator section overhead to assess the digital performance (e.g., errored seconds, severely errored seconds) and verify the connectivity, respectively. As stipulated in [8], end-to-end management of optical channels involves continuity supervision, connectivity supervision, maintenance indication, signal quality supervision, adaptation management, and protection control in the optical transmission section (OTS), OMS, and OCh layers of the optical transport network (OTN). To support these functions, the International Telecommunication Union — Telecommunication Standardization Sector (ITU-T) decided to transport the OTS and OMS overhead along the optical supervisory channel (OSC). The implementation of the OCh overhead is to date still unresolved. However, there is a consensus in preparation that utilizes the channel-nonassociated OSC in combination with a channel-associated digital wrapper around the OCh payload [9]. Awaiting the availability of these OTN management functions, GTS implemented an OCh customer impact analysis application for its lambda service [10]. Although this solution successfully resolves a number of the critical management requirements mentioned above, GTS believes that the OTN management functions will be the key enabler of future optical networking.

CONCLUSIONS WDM technology is being massively deployed in long-haul networks, such as the GTS pan-European transport network, and is now also being introduced in short-haul networks. We have shown that WDM has a considerable cost advantage, especially in long-haul networks that experience fiber exhaustion or have high fiber lease

120

or build costs. The cost savings are mainly due to reductions in SDH regenerators and transit SDH equipment (i.e., pass-through wavelengths). WDM technology not only increases transport capacity, it also supports architectural concepts like SDH and IP overlay networks and emerging native optical services. The design of an SDH-over-WDM transport network involves survivability choice, ring selection, traffic routing, and overlay ring optimization. We have shown that for GTS a network design based on interconnected MS-SPRings is only 10 percent cheaper than using SNC-P rings, and that 33 percent of ADMs could be removed and replaced by back-to-back transponders. Some design concepts for IP-over-WDM networks are provided as well. In the future the WDM network will not only consist of point-to-point WDM systems, but also of OADMs and OXCs. Today’s OADMs, however, have limited and fixed add/drop capabilities, which makes them unattractive for a new carrier. Opaque OXCs are a pragmatic and viable alternative for all-optical OXCs, especially in the short term. Reconfigurable OADMs and OXCs may reshape future provisioning and survivability processes in transport networks. In the coming years, the WDM NMS must also evolve to support the upcoming advanced networking requirements. Therefore, the network and service management layers as well as the OTN management functions urgently need to address the critical fault and performance requirements.

ACKNOWLEDGMENTS The authors would like to thank Mr. Paolo Casaschi and Dr. Bruno Meuris, both managers of Network R&D and Integration, GTS Network Services, for their valuable comments.

REFERENCES [1] J. P. Ryan, “WDM: North American Deployment Trends,” IEEE Commun. Mag., vol. 36, Feb. 1998, pp. 40–44. [2] N. Wauters et al., “Survivability in a New Pan-European Carriers’ Carrier Network Based on WDM and SDH Technology: Current Implementation and Future Requirements,” IEEE Commun. Mag., vol. 37, Aug. 1999, pp. 63–69. [3] T.-H. Wu and N. Yoshikoi, ATM Transport and Network Integrity, Academic Press. [4] P. Arijs and P. Demeester, “Efficient Design of Stacked Multiplex Section Shared Protection Rings,” Proc. NOC ’98, vol. 2, Manchester, U.K., June 23–25, 1998. [5] E. Goldstein et al., “National-Scale Networks Likely to Be Opaque,” Lightwave Xtra!, Feb. 1998. [6] D. O. Awduche et al., “Multi-Protocol Lambda Switching: Combining MPLS Traffic Engineering Control with Optical Crossconnects,” IETF Internet draft, Nov. 1999. [7] N. Lemaitre et al., “Implementation of Optical Protection in the Largest European DWDM Network,” Proc. NFOEC ’99, Chicago, IL, Sept. 26–30, 1999. [8] ITU-T Rec. G.872, “Architecture of Optical Transport Networks,” Oct. 1998. [9] G. Newsome and P. Bonenfant, “A Proposal for Providing Channel-Associated Optical Channel Overhead in the OTN,” ANSI T1 contrib. T1X1.5/99-002R1, May 1999. [10] B. Meuris and L. Missa, “The Challenge of Managing the Most Advanced Pan-European Optical Network: Strategy and Realization, A Case Study,” Proc. TeleManagement World, Las Vegas, NV, Dec. 6–9, 1999.

BIOGRAPHIES KRIS STRUYVE ([email protected]) received his M.Sc. degree in electrical engineering in 1992 and his Ph.D. degree in 1998, both from Ghent University, Belgium.

IEEE Communications Magazine • March 2000

From 1992 to 1998 he worked for the Flemish research institute IMEC on the European research projects RACE II IMMUNE and ACTS PANEL, studying network survivability. Since February 1998 he has been a member of the Network R&D and Integration department in GTS Network Services. He is currently heading the Network Integration team responsible for integrated network architecture and modeling strategy in GTS Network Services. N ICO W AUTERS ([email protected]) received a Ph.D. degree in telecommunication engineering at Ghent University in 1997 on WDM transport network architecture. During this period he worked for IMEC, a Flemish research institute, in various RACE and ACTS projects on multiwavelength optical transmission and optical networking techniques. At the beginning of 1997 he joined GTS Network Services in the R&D department, where he was responsible for developing the medium- and long-term network architecture strategy. Currently he heads up the design team of the GTS transmission network. P E D R O F A L C A O F O N S E C A ([email protected]) received in 1985 a telecommunication engineering degree from the University of Aveiro, Portugal. He is presently the Network R&D and Integration Director at GTS Network Services, where he is responsible for GTS network strategic development and planning. He started working in 1985 at Portugal Telecom/CET Research as an ASIC and system design engineer. Involved in the standardization of SDH since the beginning, he contributed to some of the ETSI technical reports and standards on SDH architecture, survivability, and network interconnection. He also participated in several Eurescom Transport Network related projects and studies. He was responsible for the first SDH tender in PT, and reached head of the SDH department in 1994. He joined Hermes Europe Railtel in 1997 as network R&D director, responsible for the introduction of DWDM and IP/WDM in the Hermes Europe Railtel network. PETER ARIJS ([email protected]) received his M.Sc. degree in electrical engineering in 1996 at Ghent University, Belgium. Afterward he joined the Department of Information Technology (INTEC) at the University of Gent, where he is working toward a Ph.D. in the broadband communication networks group headed by Prof. Piet Demeester. He worked on several projects, studying tech-

IEEE Communications Magazine • March 2000

nologies and architectures for SDH and WDM transport networks and optical access networks, and developing planning and evaluation methods for these networks. D IDIER C OLLE ([email protected]) received his M.Sc. degree in electrical engineering in 1997 at Ghent University, Belgium. Since then he has been working as a Ph.D. student at the Department of Information Technology (INTEC) at Ghent University, within the broadband communication networks group headed by Prof. Piet Demeester. His main research interests are the study of architectures for SDH transport networks and recently also packet-based networks, including development and evaluation of planning methods for such networks. PIET DEMEESTER ([email protected]) received his Ph.D. degree from Ghent University, Belgium, Department of Information Technology (INTEC) in 1988. He became a professor at Ghent University, where he is responsible for research on communication networks (supervising about 25 engineers). He was involved in over 15 European ESPRIT, RACE, and ACTS projects. His current interests are related to broadband communication networks (IP, ATM, SDH, WDM, access), and include network planning, network and service management, telecom software, internetworking, network protocols, QoS, and so on. He has published over 300 papers in the fields of optoelectronics and communication networks. PAUL LAGASSE [M] ([email protected]) received his M.Sc. degree in electrical engineering in 1969 and his Ph.D. degree in 1972, both from Ghent University, Belgium. In 1981 he became a professor of electrical engineering at Ghent University, where he is now head of the Department of Information Technology (INTEC). In 1985 he also became director of the INTEC division of the Inter-University Micro Electronics Centre (IMEC) in Leuven, and since 1993 he is Secretary General of the International Union of Radio Science. After originally working in the area of surface acoustic waves, he is now mainly active in the fields of opto-electronics, high frequency technology, and broadband telecommunications. He is a member of the board of the Flemish Institute for Science and Technology, a corresponding member of the Belgian Royal Academy of Science, and since 1997 a member of the board of governors of IEEE LEOS.

Present operational WDM networks based on point-to-point systems reveal fault and performance management requirements that are of paramount importance for the future of optical networking.

121