700 DEPLOYMENT CONSIDERATIONS 13.2.9 These networks are sometimes called storage-area networks. This is the primary application for most of the WDM networks deployed in metro networks today. Because of the nature of the traffic and a large amount of passthrough traffic in these networks, a strong case can be made for deploying WDM rings with optical add/drop multiplexers instead of higher-speed TDM rings. We present a detailed case study of a metro access network in Section 13.2.9. It is important to realize that despite the shorter spans for metro networks, optical amplifiers may still be needed for several reasons: 1. Although spans are short, in many cases the fiber in the ground is old, has many connectors in its path, and thus has relatively high loss. For example, a 10 km metro link may have a loss as high as 10 dB. 2. The loss is not just due to spansma large component of the loss comes from the loss of optical add/drop multiplexers, each of which can add several decibels of lOSS. 3. Finally, protection requirements drive the need for alternate spans that may be much longer (for example, around a ring) than the working spans. As of this writing, there has been widespread deployment of private WDM links for enterprise applications in the metro network. Several carriers in the United States have deployed WDM in their metro networks, but many are still considering the relative benefits of WDM versus other alternatives in this part of the network. As such the deployment is not yet as ubiquitous as it is in the long-haul network. Metro Ring Case Study We now look at a detailed example of upgrading a metro ring, based on a study done in [GR99]. Consider a four-node access ring with three remote nodes homing into a hub node. Assume for simplicity that all traffic is between the hub node and the remote nodes, with no traffic between the remote nodes themselves. Initially we have a SONET ring operating at OC-3 (155 Mb/s) capacity. Suppose the capacity on this ring is exhausted and that no spare fibers are available along the ring. We now have a couple of different options for upgrading the ring. The first option is to upgrade the ring to the next higher speedmOC-12 (622 Mb/s). This requires replacing or upgrading the SONET add/drop multiplexers (ADMs) at all the nodes. This is the TDM upgrade path. The other alternative is to introduce WDM and build multiple "virtual" rings at different wavelengths over the same fiber pair. We can do this in incremental steps, one additional ring at a time. For example, as shown in Figure 13.15, we can start by adding another ring at a different wavelength connecting one of the remote nodes (the one that needs more capacity, say, node 1) 13.2 Designing the Transmission Layer 701 Figure 13.15 Using WDM to upgrade a four-node access ring. One additional ring is added at a different wavelength. (a) The physical topology and (b) the lightpath topology showing the connectivity between the SONET ADMs. to the hub. In order to do this, we would need to introduce wavelength add/drop multiplexers (OADMs) at each node to drop the appropriate wavelengths. These can be "coarse" OADMs, since it is likely that the original ring is operating at 1310 nm, and we would add new rings in the 1550 nm WDM window. We would also need to add additional SONET ADMs at node 1 and at the hub, say, at OC-3 rates, if node 1 desires another OC-3 of capacity into the hub. Note that only two SONET OC-3 ADMs need to be added in this scenario. We can continue this upgrade path by adding additional rings, as shown in Figure 13.16. As we add additional rings, we will need to deploy additional "dense" OADMs at the nodes to separate out the different wavelengths used inside the 1550 nm wavelength window. The key point to note in the WDM scenario is that, compared to the TDM sce- nario, the existing SONET equipment is preserved, and additional (SONET) hard- ware is only added at nodes that need additional capacity, requiring a potentially smaller up-front capital expense. Note that a similar upgrade process can be used to upgrade an OC-12 ring to an OC-48 ring, or an OC-48 ring to an OC-192 ring. Moreover the WDM approach allows flexibility in dealing with non-SONET protocols on the different wavelengths and can provide future scalability. For example, once an OC-3 ring is upgraded to an OC-12 ring using TDM, what happens if the OC-12 ring runs out of capacity? The 702 DEPLOYMENT CONSIDERATIONS Figure 13.16 Continuing the upgrade process from Figure 13.15. Two additional rings are added at different wavelengths to the base configuration. (a) The physical topology and (b) the lightpath topology showing the connectivity between the SONET ADMs. OC-12 hardware would then have to be replaced by OC-48 hardware. In contrast, with the WDM upgrade path, additional wavelengths can be added at higher bit rates (for example, we could retain, say, two existing OC-12 rings, and add a third ring at OC-48 or OC-192). Therefore the WDM solution is more "future-proof," compared to the TDM solution. The key question we've left unanswered is how the two approaches compare from a cost perspective. This depends to a large extent on the cost of the OADMs relative to the SONET ADMs. Figure 13.17 shows the network cost for the upgrades described in this example, assuming the equipment costs shown in Table 13.6. The three sets of lines correspond to an upgrade from an OC-3 to an OC-12 ring, an OC-12 to an OC-48 ring, and an OC-48 to an OC-192 ring. For the cost numbers we have assumed, it appears that at low bit rates (OC-3 to OC-12) it is more cost-effective to do a TDM upgrade, whereas at the higher bit rates (OC-48 to OC-192), it may be cheaper to do a WDM upgrade. These numbers clearly will change over time based on the relative costs of the OADMs and the SONET ADMs, but the trends shown in Figure 13.17 will still be the same. We leave it to the reader to generate the numbers used to plot the lines in Figure 13.17. 13.2 Designing the Transmission Layer 703 r~ r~ 0 r/3 0 0 0 0 z $8OO $700 $600 $500 $4OO $300 $200 $100 $0 TDM OC-48 to OC- 192 u OC- 12 to OC-48 upgrade OC-3 to JAW OC-12 upgrade DM ~/WDM A A A i" ~j " TDM 100 1000 10,000 Network capacity (Mb/s) Figure 13.17 Relative network cost for the TDM and WDM upgrades. Three sets of upgrades are shown: OC-3 to OC-12, OC-12 to OC-48, and OC-48 to OC-192. The horizontal lines indicate the TDM upgrade path, and the slanted lines indicate the WDM upgrade path. Table 13.6 Equipment cost assumptions for Figure 13.17. Prices of all the equipment listed here are coming down, due to competition and improvements in technology. These numbers are reasonably indicative of prices in 1999/2000. Equipment Cost without WDM Interfaces (U.S. $) Cost with WDM Interfaces (U.S. $) OC-3 ADM OC-12 ADM OC-48 ADM OC-192 ADM Coarse OADM Dense OADM 25,000 60,000 175,000 15,000 35,000 80,000 1 O, 000 20,000 704 DEPLOYMENT CONSIDERATIONS 13.2.10 From Opaque Links to Agile All-Optical Networks The optical layer itself is evolving, not just in terms of raw capacity, but also in terms of functionality. The optical network originally consisted of WDM links, with all the functions at the end of the link performed in the electrical domain. These networks are sometimes called opaque networks. Due to the high cost of optical-to-electrical (O/E) conversions, particularly at the higher bit rates, it makes sense to minimize the number of these converters in the network. The first step in this direction was the development of ultra-long-haul systems, which provided longer reach between regenerators. The second step is to handle as much of the traffic passing through a node in the optical domain as possible. An all-optical OADM or OXC performs this function. Having optical passthrough instead of electrical processing can lead to an order of magnitude savings in the cost, given that the cost of O/E conversions dominates the cost of the node itself. There are associated savings in power and floor space as well, given that the O/E devices consume most of the power and occupy most of the floor space in WDM equipment. Even further cost savings can be realized by passing signals through in bands of wavelengths, instead of individual wavelengths. These networks are called all-optical or transparent networks. The next step in the evolution of the optical layer is to add agility. An agile network provides the ability to set up and take down lightpaths as needed and allows carriers to provision and deploy services rapidly. With the introduction of optical crossconnects and reconfigurable optical add/drop multiplexers, agile opaque networks are becoming a reality today. It is only a matter of time before agile all-optical networks arrive. Adding agility to an all-optical network results in even further complexities to be tackled at the physical layer, such as adaptive power and dispersion management. These problems have already been tackled at least partially in the context of ultra-long-haul systems. While an all-optical network provides significant advantages, it also has its limi- tations. Certain functions, such as wavelength conversion, regeneration, and traffic grooming at fine granularities (for example, at STS-1 or 51 Mb/s) will need to be done in the electrical domain. As we saw in Chapter 8, we may not be able to com- pletely handle all the passthrough traffic in the optical domain, due to inefficiencies in how traffic is groomed in the network. For these reasons, a practical node will end up using a combination of all-optical and electrical crossconnects. The all-optical crossconnects can be used to switch signals through in the optical domain as much as possible, and signals needing to be regenerated, converted from one wavelength to another, or groomed will be handed down to the electrical layer. Another subtle aspect of the all-optical network is related to interoperability between systems from multiple vendors. As we saw in Chapter 9, it is difficult for Summary 705 equipment from different vendors to interoperate at the wavelength layer. Interop- erability between vendors needs to be done through regenerators/transponders. This implies that the all-optical network by itself is a single-vendor network. Transpon- ders are needed at the edges of this network to provide interoperability with other all-optical networks. A realistic network will therefore consist of all-optical islands or subnets, interconnected with other such subnets through transponders at the boundaries. Summary This chapter addressed architectural alternatives for the new generation of carrier networks. These networks are different from the established legacy networks based on SONET/SDH. This is driven by the increasing dominance of data over voice and the emergence of new carriers with vastly different business models offering different types of services. An established carrier offering a mix of services may choose to overlay SONET/SDH and IP or ATM over the optical layer. New carriers offering predominantly data-oriented services may opt to deploy IP or ATM directly over the optical layer, not deploying any SONET/SDH at all. The optical layer is becoming ubiquitous in both long-haul and metro networks. The optical layer here provides circuit-switched lightpaths to the higher layers. Note that the optical layer is not performing any packet-switching functions. These functions are best left to the electronic layers. Optical packet-switching technology is still in research laboratories. The next-generation metro access network will likely use a hybrid packet-circuit network element as the key element to deliver services. The core of the network is mi- grating away from a SONET ring-based architecture to a meshed optical-layer-based architecture, with protection functions implemented in the optical layer. Within the optical layer, TDM, WDM, and SDM are all used to provide capacity. The right combination of these techniques is not an easy choice and depends on a variety of factors including the length of the link, the availability of spare fibers, the type of fiber and its dispersion and nonlinear characteristics, and the type of services to be deployed using the network. The problems at the end of this chapter will give the reader an inkling of what such a comparison might involve. Network planners need to make their own analysis of the different alternatives, perhaps with the aid of some network planning and design tools, to decide which way to go. The optical layer itself is migrating from an opaque network, consisting of WDM links with electrical processing at the ends of the link, to an all-optical network, where traffic is passed through in the optical domain at intermediate nodes. At the same time, the optical network is moving from a static network to an agile network, where lightpaths can be set up and taken down as needed. 706 DEPLOYMENT CONSIDERATIONS Further Reading The subject matter in this chapter is widely covered in the business press and by investment houses. Several market research firms publish reports on vari- ous segments of the optical networking industry. These include Communications Industry Researchers (www.cir-inc.com), Electronicast (www.electronicast.com), KMI (www.kmicorp.com), Ovum (www.ovum.com), Pioneer Consulting (www. pioneerconsulting.com), Ryan, Hankin and Kent (www.rhk.com), Strategies Unlim- ited (www.strategies-u.com), and Yankee Group (www.yankeegroup.com). There have been many studies published about the relative economics of various architec- tural options. Be warned these are rather biased views, as the assumptions made significantly impact the outcome, and these assumptions are usually biased toward supporting the products offered by the vendor doing the study. The various op- tions for supporting IP over WDM have been explored in many papers; for instance, [Mae00] provides a relative cost analysis. See [PCW+00, Coo00, OSF00, PCK00] for a recent sampling of papers related to metro WDM economics. [DSGW00, Dos01] explore the value proposition behind ultra-long-haul WDM systems. The National Fiber Optic Engineers' Conference usually has many papers on these topics. 13.1 Problems Imagine that you are a planner for a long-haul carrier planning to deploy an IP over WDM network. Your job is to make the right technology and vendor choice for your network. You are given the following information. The initial requirement is to deploy 20 Gb/s of capacity between two nodes. You anticipate that this capacity will grow to 80 Gb/s in a year and over a few years grow to 320 Gb/s. You have a choice of several WDM systems from different vendors with the following prices and capabilities: Vendor A B C Number of channels 80 128 32 Bit rate per channel OC-192 OC-48 OC-192 Distance between regenerators 640 km 1920 km 1920 km Amplifier spacing 80 km 80 km 80 km OLT common equipment $200,000 $275,000 $300,000 Transponder $50,000 $25,000 $80,000 Amplifier $150,000 $100,000 $125,000 Problems 707 13.2 13.3 Assume that the common equipment prices for the optical line terminals include any amplifiers if needed. One transponder is needed for each channel at each end of the link. Once the distance between regenerators is exceeded, the signals need to be regenerated by using two terminals back to back with transponders. Compute the cost of each solution for a 640 km link, a 1280 km link, and a 1920 km link. Draw a diagram of each configuration. What are your conclusions? Other than the costs computed above, what other factors might influence your choice? Consider the same problem as in Problem 13.1 with one difference. For the 1280 and 1920 km cases, between the two nodes is a third node spaced 600 km from the first node, where half the capacity needs to be dropped and added. For this case, assume that vendor B and vendor C offer systems where you can use back-to-back terminals at this intermediate node without requiring transponders for the passthrough chan- nels. (Transponders are still needed for the channels dropped and added.) Repeat your analysis. What are your conclusions? Imagine that you are a planner for a metro carrier. The links in your network are fairly short, with a maximum span length of 40 km. You want to compare SDM, TDM, and WDM options for realizing a two-node link. Assume the following costs. Equipment Cost (U.S. $) Pulling fiber through existing conduit (per km) Laying new conduit, including fiber (per kin) OC-48 BLSR/2 ADM Common equipment Additional per OC-12 drop Additional per STS-1 drop OC-192 BLSR/2 ADM Common equipment Additional per OC-12 drop OC-12 BLSR/2 ADM Additional per STS-1 drop Metro WDM terminal (OLT) Common equipment Additional per transponder 300 20,000 40,000 5,000 750 125,000 5000 15,000 750 30,000 10,000 708 DEPLOYMENT CONSIDERATIONS You need to deliver 10 Gb/s of capacity in the form of OC-12s (622 Mb/s) to your customers. Compare the cost of the following options for the scenario where fibers are available versus fiber needs to be pulled through existing conduit versus new conduit needs to be laid: (a) OC-48 ADMs over separate fibers, (b) OC-48 ADMs in conjunction with WDM terminals over a single fiber pair, (c) OC-192 ADMs, and (d) WDM terminals with no SONET equipment. Factor in the cost of protection as well. Assume that two diversely routed fiber pairs are available between the two sites. Whenever SONET is used, protection is done in the SONET boxes, and no protection is done in the OLTs. For the case with no SONET equipment, the protection is done at the optical multiplex section by the OLTs~assume that the cost is already factored into the OLTs. Repeat this problem for the case where the capacity needs to be delivered in the form of STS-ls to your customers. In this case, the options available to you are: (a) OC-48 ADMs over separate fibers, (b) OC-48 ADMs in conjunction with WDM terminals over a single fiber pair, (c) OC-192 ADMs back-ended by OC-12 ADMs, and (d) WDM terminals back-ended by OC-12 ADMs. Draw a diagram of the different configurations. What are your conclusions? Other than the costs computed above, what other factors might influence your choice? 13.4 This is an extension of the previous problem related to planning a metro network. We will explore the use of optical add/drops in this problem. You now have to create a linear network of three nodes A, B, and C. The link between node A and node B is 40 km, and the link between node B and node C is also 40 km. You need 5 Gb/s of capacity between A and B, 5 Gb/s between B and C, and another 5 Gb/s between A and C. All capacity is to be delivered as OC-12s. In addition to the equipment available above, you also have the option of using an OADM at node B that works with OLTs at node A and node C. The WDM sys- tem has a reach of 80 km with an intermediate OADM. The OADM has a com- mon equipment cost, including any needed amplifiers, of $50,000 and can drop as many wavelengths as needed. Transponders are needed for the added and dropped channels. In addition, assume that SONET ADMs have a maximum reach of 40 km. Signals need to be regenerated after this, and the regenerator costs are as follows: OC-48 regenerator, $10,000; OC-192 regenerator, $30,000. Now consider the following solutions: (a) Fibers are available, and you use OC-48 ADMs over them. In this case you need to use a regenerator at node B for passthrough traffic or another OC-48 ADM for multiplexing and demultiplexing local traffic. Consider also the cases where fiber needs to be pulled through existing conduit and also of conduit exhaust. (b) OC-48 ADMs along with OLTs and OADMs. Problems 709 13.5 (c) OC-12 delivery directly using OLTs and OADMs, no SONET. (d) OC-192 ADM with another ADM at node B to demultiplex and multiplex local traffic. For this problem, ignore any protection needed. Note that this could result in cheaper equipment, but for our purposes, assume that the equipment costs don't change. Compare the costs of these alternatives. What do you conclude? You are looking at deploying an optical crossconnect at a large node in a carrier net- work. The crossconnect is connected to OLTs and drops traffic down to IP routers. You have three options to consider: (1) an electrical crossconnect (EXC) solution, where the crossconnect uses short-reach interfaces connected to transponders in the OLTs and to short-reach interfaces in the routers; (2) an opaque photonic crosscon- nect solution, where the photonic crossconnect (PXC) is connected to transponders in the OLTs and to short-reach interfaces in the routers; (3) a transparent photonic crossconnect solution, where the photonic crossconnect is connected to the OLTs directly without transponders, but transponders are used between the routers and the crossconnect. Assume the following: Item Cost (U.S. $) Power Footprint WDM OC-48 transponder $25,000 75 W 64 ports/rack WDM OC-192 transponder $50,000 150 W 32 ports/rack EXC switch fabric 10,000 W 1 rack EXC OC-48 port $15,000 50 W 256 ports/rack EXC OC-192 port $30,000 100 W 64 ports/rack PXC port $25,000 2 W 256 ports/rack Assume that the EXC supports a maximum of 512 OC-48 ports or 128 OC-192 ports and that the PXC supports 1024 ports. Compare the cost and floor space taken up for the three options above for the following situations. (Include any transponders used, but neglect the routers as they are common to all the scenarios.) Summarize your findings. (a) The node is switching 256 OC-48 wavelengths coming in from the WDM systems, of which 25%, 50%, or 75% of the traffic may be dropped locally into router ports. (For example, with a 25% drop, you would need a total of 320 ports on the crossconnect.) (b) The node is switching 256 OC-192 wavelengths coming in from the WDM systems, of which 25%, 50%, or 75% of the traffic may be dropped locally into router ports. . problems have already been tackled at least partially in the context of ultra-long-haul systems. While an all -optical network provides significant advantages, it also has its limi- tations. Certain. to an upgrade from an OC-3 to an OC-12 ring, an OC-12 to an OC-48 ring, and an OC-48 to an OC-192 ring. For the cost numbers we have assumed, it appears that at low bit rates (OC-3 to OC-12). WDM terminals over a single fiber pair, (c) OC-192 ADMs back-ended by OC-12 ADMs, and (d) WDM terminals back-ended by OC-12 ADMs. Draw a diagram of the different configurations. What are your