A few years ago, it wasn't cool to say Sonet was sexy. Sonet was considered incredibly inefficient and inflexible – a stodgy, old-fashioned circuit technology that wouldn't survive in tomorrow's packet-based networks.

Now, everything's changed. The need for carriers to make a better return on investment has forced them to recognize that nearly all of their revenues come from services provided over circuits. At the same time, Sonet has undergone some radical improvements – notably ones that slash costs and eliminate many of its ineffiencies in handling packet-based traffic.

These developments have particular relevance to metro networks, where most of the action is right now, for a couple of reasons. First, this is where the rubber hits the road in terms of marshalling packet-based traffic and feeding it into the big fat optical pipes in carrier backbones. Second, cost considerations are very important in metro networks.

This report digs into these developments. It's part of a series of articles that freelance analyst Tim Hills has researched and written for Light Reading. The series is intended to put metro technologies into context, make them more understandable, and help folk see through the marketing hype pumped out by component and equipment suppliers.

The first report, Metro Multiservices Evolution, set the scene by explaining why metro networks are hot, the challenges facing carriers, and the technologies being proposed to solve these problems.

This report focuses on what's become known as Next-Generation Sonet. Its emphasis is on systems, while another report, covering Sonet chip developments, is scheduled for publication here in early June. (It will be previewed in a free, live Webinar on May 28. Click here to register.)

Subsequent reports in this series will delve into the details of "carrier-class" Ethernet, Resilient Packet Ring (RPR), and dense wavelength-division multiplexing (DWDM) developments.

This report starts by examining what the term Next-Generation Sonet really means and then reviews the technologies covered by the brand name. It then goes on to identify key trends in chip developments and outlines the ways next-gen Sonet is being built into metro equipment. Finally, it surveys a selection of products from key vendors in this space.

Here's a hyperlinked summary:

The Next-Gen Brand

Next-Generation Technologies

Chips and Devices

Using Next-Generation Sonet

System Developments

Some of this report digs quite deeply into technology issues. To get the most out of it, why not start by listening in to the archived Web-enabled preview, given by Scott Clavenna, Director of Research at Light Reading. Click here to do this.

Here's some background reading that might also help:

Introduction by Peter Heywood, Founding Editor, Light Reading
http://www.lightreading.com

About the author: Tim Hills is a freelance technical writer. He may be reached at: [email protected].

Next Page: The Next-Gen Brand

If metro network technologies were soap powders, Next-Generation Sonet would be a leading brand. But that’s part of the problem for carriers trying to figure out how next-gen Sonet fits into their metro plans – it’s just a brand, not a single technology or standard.

To help get a fix on where next-gen Sonet is today, this report reviews the technologies and industry drivers behind the brand and samples some of the products on the market under the Next-Generation Sonet banner.

Pretty well all Sonet vendors have, or will soon release, products they describe as next-generation Sonet in some sense. But it's a terminological nightmare (and a marketeer’s delight) because no one can quite agree on what it is exactly – though it hardly seems to worry the brochure writers, as the marketing usually emphasizes the packet and client services (which, admittedly, generate the revenues), not the underlying transport mechanism. This might be OK, but for the fact that the transport mechanism does matter, because it determines some of the ultimate capabilities of the carriers’ network and service offerings. Powder X may indeed not wash as clean as Powder Y in the long term.

Wavelength Division Multiplexing (WDM) makes things even more complicated, as most multiservice provisioning platforms (MSPPs – prime users of next-gen Sonet) are now going over to DWDM, and it is very unclear whether wavelength-aware digital wrapping schemes such as G.709 can be usefully regarded as descendants of Sonet or not – even though they incorporate various Sonet ideas (check out our Beginners' Guide for more on Digital Wrappers and Forward Error Correction (FEC)).

But take a look at Sonet’s fundamentals, and some clues emerge to what next-generation Sonet might mean at a technical minimum. Traditional Sonet does four basic things:

Sonet’s real weak point for today’s carrier requirements is the old-fashioned and rigid multiplexing scheme and its inefficient use of single-wavelength paths for modern packet-oriented traffic. Especially since Gigabit Ethernet has taken over from Asynchronous Transfer Mode (ATM), in terms of enterprise interest, and has become a big network requirement.

“I am simply stunned how fast the market in the metro shifted from ATM to Ethernet. Two years ago every little startup out there was doing an ATM-over-Sonet box, and now it’s almost nobody. Ethernet is really the dominant one, and it makes sense because it’s inexpensive, it’s ubiquitous, and enterprises understand it,” says Richard Goode, senior manager of the Optical Networking Group at Lucent Technologies Inc. (NYSE: LU).

What’s needed are ways to manage data-service bandwidth dynamically in small increments, to provide a range of service guarantees, and to engineer traffic flows more efficiently. So to improve Sonet into a new generation, while keeping its essential virtues, the main technological focus is on devising new client-service encapsulations and scrapping the traditional multiplexing/mapping scheme, replacing it with a more flexible alternative within the basic Sonet framing.

Bigger Picture – Costs and Densities

But, of course, there is a bigger industry picture behind next-gen Sonet than just new encapsulations and mappings. Classical Sonet has had the stigma of being expensive, difficult, and complex, with time-consuming service provisioning, although its solid OAM has been appreciated by incumbents.

The collapse in telecom finances has left the surviving carriers desperate to cut costs and to raise revenues – but in an environment where bandwidth growth has slowed, as has the premium that enterprise customers are prepared to pay for connectivity. Since the survivors tend to be big incumbents that have already invested heavily in Sonet, vendors of Sonet equipment naturally see next-generation Sonet in this light, too.

“It really takes existing Sonet – which is a standards-based technology – and makes it better through increasing the densities, reducing the footprint, and reducing the costs,” says Goode. “Currently, the carriers’ message to us is they are not asking how many technologies you can integrate into one box, but rather how can you help me save money. So they really have to do more with less. So they are still looking for the benefits of MSPPs, but those functions that the MSPP box does have to be done well.”

Continual improvements in semiconductor technology have increased densities by up to 50 percent in some cases – say, being able to support 80 OC48s in four chassis instead of 40 in three. And higher densities mean smaller footprints. Power consumption is falling, too. So reduced space leads to lower real-estate costs, and reduced power consumption leads to lower power costs – both leading to reduced operational expenses (opex).

And the savings from next-generation Sonet have to be big – and include functional and architectural gains as well. So says Joe Padgett, director of the Metro Optical Group at Nortel Networks Corp. (NYSE/Toronto: NT): “It has got to set a new benchmark in economics. So you are consolidating nodes, you are consolidating some of the rings. If you are not getting a 25 percent to 35 percent capex/opex savings, then you are definitely not next-generation Sonet.”

Evolution and Legacy Support

Evolution is another key attribute of the next-generation Sonet brand. Carriers have to be able to do new things, such as extending enterprise broadband service capabilities into a shared point-to-multipoint environment, while keeping their revenue-generating traditional services.

“The real key of next-generation Sonet is that it has got to support the legacy protocols and the legacy services,” says Padgett. "If you don’t do that, then to me you are not meeting the definition of next-generation Sonet as an evolution. Next-generation Sonet is not a forklifting. This is basically taking what you’ve got and gaining some efficiencies out of it."

Even newer carriers are having to take account of legacy considerations. Says Nazik Andonian, director of Integrated Network Design at the European carrier Interoute Telecommunications: “Our approach is to align the service requirements of our customers with the solutions offered, so we are currently studying the capabilities of next-generation SDH platforms. Interoute is penetrating the metropolitan area; and providing an efficient last-mile access solution for carrying both Ethernet and voice traffic is a significant step. Interworking with any installed-base metro platform is obviously a major concern.”

Next-Generation Challenges

There are also some specific challenges that next-gen Sonet equipment must meet. These include:

Next Page: Next-Generation Technologies

The traditional way of transporting packet data directly on Sonet without an ATM layer is by Packet-Over-Sonet (POS) protocol. POS is mature, widely available, and cheap in some configurations – but it is seen by much of the industry as a deadend as far as next-gen Sonet is concerned. POS just encapsulates frame/packet-based protocols like PPP (point-to-point protocol), Frame Relay, and Ethernet over HDLC (high-layer data link control) within Sonet; it cannot differentiate the separate packet streams on the link to allow per-stream traffic engineering, protection, and bandwidth management.

So, although POS is being used by some vendors in concatenated Sonet containers as small as STS3c, this does not provide the 1-Mbit/s Ethernet granularity that many enterprise users want. POS is basically a dumb link that relies on higher-level devices such as routers to do its traffic engineering and service creation.

POS’s importance for large concatenated pipes and as a legacy protocol means that it still has development life, however. A new chip interface called the POS-PHY Level 4 chip interface has been accepted as a standard by the Optical Internetworking Forum (OIF) and the ATM Forum to provide an optimized interface for next-generation interfaces for 10-gigabit POS, ATM, and Ethernet components.

POS-PHY Level 4 supports highly channelized applications, POS or ATM, from OC192 (10 Gbit/s) down to STS1 (51.8 Mbit/s) or 10-Gigabit Ethernet down to Fast Ethernet. Commercial chipset vendors, such as PMC-Sierra Inc. (Nasdaq: PMCS) (one of the standard’s promoters) have released devices that support OC192c and high-density OC48c POS applications in addition to ATM and 10-Gigabit Ethernet. For future 40-gigabit applications, POS-PHY Level 5, a logical extension of POS-PHY Level 4, has recently been ratified at the OIF.

POS-PHY Level 4 is aimed squarely at metro multiservice platforms as an aid to convergence of existing networks. Steve Perna, vice president and general manager of the Optical Networking Division of PMC-Sierra, observes: "POS-PHY Level 4 provides an excellent method for next-generation routers and switches to support 10-Gbit/s POS, ATM, and 10-Gigabit Ethernet."

As always, the industry has come up with several divergent routes in the standards process to overcome the limitations of POS, and perceptions play a big part. One philosophy argues that what matters is first to fix Sonet to carry Ethernet properly, as Ethernet hugely dominates user native packet traffic and is an immediate market for carriers. A more ambitious approach aims to handle all forms of native packet traffic over Sonet, so that carriers can profit from both Ethernet and new markets like storage area networks. Of course, some vendors have launched proprietary technologies with their own take on how to fix Sonet for Ethernet.

The result is a veritable Babel of protocol acronyms and codes that reflect different aims and – up to a point – different underlying technologies. The caveat is that lurking under them is the basic Sonet frame, usually hollowed out by some form of capacity concatenation to accept new client-service mappings. To confuse matters even more, vendors tend to talk of Ethernet over Sonet as if it were a single standard or technology. It isn’t.

The names to note are:

WIS

WIS (Wide-Area Interface System) is rather special and is only just entering the running, as it is part of the ongoing IEEE 802.3ae 10 Gigabit Ethernet initiative. WIS is a digital wrapper compatible with concatenated Sonet/SDH OC192c/STM64c to provide an Ethernet WAN PHY as an alternative to the LAN PHY for native-format, dark-fiber networks. Chip framers for both 10-Gigabit Ethernet and Sonet/SDH OC192c/STM64c are beginning to appear on the market – for example, from Silicon Packets, a startup acquired by Cypress Semiconductor Corp. (NYSE: CY) early this year (see Cypress Improves Its Line Card Lineup) – in time for the expected completion of 10-Gigabit Ethernet standards by 2003.

X.86

X.86 is Ethernet over Link Access Procedure – SDH (LAPS), approved by the International Telecommunication Union, Standardization Sector (ITU-T) in February 2001. In its current form, X.86 is a simple method of providing Ethernet LAN extension over a public or private WAN, allowing 802.3 Ethernet switches and hubs to interface directly with an SDH network for point-to-point communications. It offers low latency variation, remote performance monitoring, remote fault indication, and active flow control of burst traffic. So carriers can use X.86 for Ethernet private-line services with end-to-end traffic segregation, security, interoperability, and guaranteed service rates.

LAPS is a connectionless HDLC-like framing structure that encapsulates 802.3 Ethernet MAC frames; it derives from the earlier X.85 standard for IP over LAPS, where a big point was to get rid of PPP from the PPP/HDLC combination of POS. X.86 does not specify how LAPS is actually mapped into SDH, but it works with a wide range of concatenated and nonconcatenated payloads – VC, VCc, STM-N, and sSTM-N. It is basically a rate-adapting bridge layer between Ethernet and SDH.

Some vendors already support X.86. According to P. G. Menon, cofounder and VP of marketing at Atoga Systems, a big attraction is that X.86 allows individual Ethernet traffic streams to be switched and groomed, so that link capacity is not wasted, as often happens when streams are multiplexed into fixed Sonet pipes. So carriers can build lower-cost and more manageable converged networks, with Sonet capacity split between Ethernet and TDM services.

“A 100-Mbit/s flow mapped into an STS3 achieves a 66 percent utilization at the STS level,” he says. “If each flow is mapped into multiple STS1s in increments of 64 kbit/s, you achieve a 99 percent utilization.“

However, at least in the U.S., there appears to be no widescale rush by carriers to jump into X.86. Says Lucent's Goode, “There are several startups using X.86, but only one of the major ILECs over here has advocated the use of X.86 from a technology perspective. Everyone else, we think, is migrating to the Generic Framing Procedure strategy.”

GFP/G.7041

Technically, there may not be a lot to choose from between X.86 and GFP for straight Ethernet over Sonet, but industry momentum seems to be growing behind GFP, which was recently approved by the ITU-T as G.7041 (see IEEE and ITU Get Chummy on RPR). GFP has been brewing for some years, and forms are already embedded in some vendors’ products, with the first commercial GFP chipsets now reaching the market (see Cypress Preps Ethernet-Over-Sonet Chip, Agilent Boosts Ethernet-Over-Sonet and PMC Pushes Sonet Silicon).

GFP does have some features that X.86 lacks, such as header error correction and channel identifiers for port multiplexing.“The channel identifier means that I can multiplex multiple physical ports into one path through the network. For example, TranSwitch is developing a chip that has eight 10/100 Ethernet interfaces. All eight interfaces could share one GFP path, and the channel identifier would indicate which one has frames going over that channel,” says Bill Bartholomay, chief technology officer at TranSwitch Corp. (Nasdaq: TXCC). “LAPS [in X.86] is one-to-one – one port versus one path.”

The crucial point about GFP is that it has potentially broader applications than X.86 because it has two modes of operation: frame mapped and transparent. Frame-mapped mode operates on data signals that have been packetized/framed at Layer 2 or higher by the client service feeding the Sonet network, mapping one client packet/frame into a GFP frame. X.86 works on a similar principle. Frame mode supports rate adaptation and multiplexing at the packet level for traffic engineering and aggregation at STS and VT (virtual tributary)levels. Transparent mode is completely different, because it accepts native block-mode data signals and uses the Sonet frame merely as a lightweight (low-overhead, low-latency) digital wrapper. So, in principle, GFP should be able to handle fairly well anything a carrier throws at it.

GFP’s protagonists argue that this gives carriers an easy migration to a protocol-agnostic transport layer or optical convergence layer – something for which Layer 2 protocols like Ethernet are not well-suited – as well as providing the flexibility and granularity required for Ethernet services.

Says Steve Duffy, manager of Strategic Marketing and Solutions at ONI Systems Inc. (Nasdaq: ONIS): “The reality is that there are really four major services that are going to make money for carriers – and those are not just Ethernet. Storage is just as important and deserves to be carried in its native format, which is Fibre Channel or Escon; wavelength services are also interesting; and a lot of the market is still TDM data private-line circuits. Looking at an Ethernet-only network design prevents carriers from harnessing the entire market opportunity.”

VC

Two key technologies that enable next-generation Sonet/SDH are now coming to market – Virtual Concatenation and its associated Link Capacity Adjustment Scheme (LCAS). Standards have emerged from the ITU-T and American National Standards Institute (ANSI) over the last couple of years, and some vendors will undoubtedly implement them in favor of proprietary concatenation schemes, especially to support the ITU-T G.7041 Generic Framing Procedure.

“ONI agnostically supports both Sonet/SDH and WDM transport schemes. Providing the ability for carriers to leverage Virtual Concatenation is very important, for it further improves the data-carrying efficiencies of the Sonet/SDH networks,” says Duffy. “For example, carriers can map a rate-shaped Ethernet flow of any arbitrary bandwidth to a corresponding and appropriate number of VT1.5 or STS1 channels. The benefits are that little bandwidth is wasted and you now have a more efficient scenario for carrying data over the Sonet/SDH network.”

The big ideas of VC and LCAS are to create fine-tuned and variable capacity Sonet/SDH pipes to match the needs of packet data QOS (quality of service) and customer SLAs (service-level agreements) – and to boost carriers’ traffic-handling scaleability and efficiency. A further attraction of Virtual Concatenation is that it works across suitable legacy Sonet/SDH networks, because only the end-of-pipe nodes need to be aware that VC is being used.

Sonet/SDH concatenation is not new and is provided for in the original standards. Contiguous concatenation is the simpler, but less flexible, form. This in effect merges adjacent lower-order payload containers carried within a single OCn/STMn signal into ones of greater capacity, the idea being to allow Sonet/SDH to carry new services outside the then prevalent PDH hierarchy.

Unfortunately, contiguous concatenation gives very big jumps in capacity. If you want a 160-Mbit/s pipe for your Ethernet connection, it won’t fit into a standard STS3c/VC-4 payload, so the carrier will have to use a 600-Mbit/s STS12c/VC-4-4c contiguous payload in an OC12/STM4 signal – and the rest of the capacity is wasted. In addition, all intervening network elements (for example, crossconnects) between source and destination must be capable of supporting the contiguous payload structure, something that cannot be taken for granted in a multi-domain network, particularly in the SDH environment.

Virtual Concatenation (VCX) is different. In effect, the client service is divided among several smaller payload containers that are individually transported and then recombined at the destination node.

Excess capacity is not wasted, because jumps can be much smaller (for example, in increments of a 48-Mbit/s STS1/VC-3, from 48 Mbit/s to 10 Gbit/s in principle). So a virtually-concatenated OC12/STM4 signal would allocate 4xSTS1/VC-3 (denoted as STS1-4v or VC-3-4v) for your 160-Mbit/s pipe and could use the remaining 8xSTS1/VC-3 for other traffic. Not only does this make the network more scaleable, it also uses bandwidth more efficiently and reduces the amount of stranded bandwidth throughout a carrier’s network. Within a VCX link, the individual STS/VC constituents appear to be no different from standard STS/VC payloads and so are compatible with all and any existing Sonet/SDH network elements.

Virtual Concatenation can also allow new, efficient, shared protection mechanisms, since traffic can be sliced into several parts and sent by different paths. "I think that this type of VCX capability will turn into a major benefit for network operators," says Mark Lum, portfolio solutions manager for Optical Metro at Nortel. "If you do it properly, you can dynamically adapt your service bandwidth on the fly and also provide new soft protection schemes – and so achieve higher transport network efficiency. New standards such as ITU-T G.7042 (Link Capacity Adjustment Scheme) provide basic mechanisms that enable these new possibilities."

These capabilities make Virtual Concatenation more technically complex. It concatenates payloads across multiple STS/STMs by byte interleaving them, and the different STS/STMs may follow different paths. This will lead to differential delays at the receiving end, and mechanisms are needed to correct these and to reassemble the payload. Fortunately, this has been studied extensively during the standardization process, and practical implementations will ensure that this is not a network concern.

LCAS

LCAS (G.7042, approved by the ITU-T in November 2001) builds on Virtual Concatenation by allowing the carrier to adjust the pipe capacity while it is in use. It is a two-way signaling protocol that runs continuously between the two ends of the pipe and ensures that commands from the network management system to alter the pipe capacity do not impair the user’s traffic. LCAS is a highly significant addition to Sonet’s capabilities.

Says Andrew Schmitt, Sonet/SDH product line director for Vitesse Semiconductor Corp. (Nasdaq: VTSS), “The Layer 3 IP guys are claiming they can eliminate Sonet by absorbing the transport functions. I would argue that Virtual Concatenation, GFP, and LCAS allow Sonet to cheaply move into the Layer 2 switching domain.”

GMPLS

GMPLS is not itself a Sonet technology, but it will play a huge role in networks that use next-gen Sonet because it provides a universal scheme of automatic end-to-end bandwidth provisioning across TDM, packet, and wavelength services and networks.

“This one is critical because it helps to lower the costs of delivering existing services such as voice and private lines, something which carriers have given more priority than delivering new services in today's environment,” says Kevin Wade, spokesman for Turin Networks Inc.

GMPLS is an evolving technology, still incompletely standardized. It will be covered in more detail in the forthcoming Light Reading report on metro DWDM. Past reports covering GMPLS include Tutorial on Grooming Switches and Optical Signaling Systems.

RPR

Resilient Packet Ring turns a Sonet ring into a distributed packet switch to optimize bandwidth use. It will be the subject of the upcoming Light Reading report on metro RPR.

Subcarrier Sonet multiplexing

Virtual Concatenation does have its critics. Sorrento Networks Corp. (Nasdaq: FIBR), for example, believes this approach is not necessarily optimal for obtaining high-density transport, arguing that combining "thin-mux" time-division based schemes with DWDM can be much more advantageous and cost-effective.

Sorrento uses a variety of such solutions, both asynchronous (self-timed) and synchronous (i.e., Sonet/SDH) variants, to provide a full range of "sub-rate" multiplexing options. For example, the company’s asynchronous Electronic Photonic Concentration (EPC) system can flexibly multiplex up to eight 125- to 200-Mbit/s native-format duplex client signals into a wavelength to optimize DWDM capacity use. Synchronous multiplexers aggregate smaller TDM and data clients into Sonet/SDH-framed wavelengths (2.5 or 10 Gbit/s), facilitating direct legacy handoffs and retaining TDM monitoring capabilities – a key advantage, according to the company.

“What we are focusing on is solving a problem, and not trying to do everything,” says Dr. Demetri Elias, Sorrento's VP of marketing. “We have a high-density system that provides robust DWDM transport over multiple topologies – point-to-point, linear, ring, and even mesh. As such, our solution allows operators to extract large amounts of additional capacity – up to 64 wavelengths – from their existing fiber plants. Meanwhile, so-called next-generation Sonet boxes and/or multiservice provisioning platform solutions only offer Virtual Concatenation to improve utilization over a single channel. Generally speaking, these schemes require complex electronic hardware and yield much increased footprint/power-consumption versus our thin-mux offerings. Really, is there justification in paying more to pack a single channel tributary via Virtual Concatenation, versus paying less and exploiting the full advantages of multichannel DWDM multiplexing and edge multiplexing?”

Next Page: Chips and Devices

Next-generation Sonet critically depends on implementing new functions into silicon. Unsurprisingly, chipset vendors see lots of life in Sonet and are keeping up the development pressure.

Says PMC-Sierra's Perna: “The bulk of our investment going forward will be in Sonet-based systems. That’s what our customers are building and wanting, and that’s what the carriers are requesting. Putting new services over Sonet is going to be a new area of emphasis.”

He sees a growing industry migration away from ASICs (application-specific integrated circuits) to standard chips for implementing Sonet. “There are growth opportunities remaining in this part of the business. We want to be a total solutions provider for 2.4-, 10-, and, in the future, 40-Gbit/s metro solutions that take advantage of the Sonet/SDH network infrastructure, and then provide enabling solutions such as Ethernet over Sonet and pure packet networks when they evolve.”

The commercial chip vendors are driving their Sonet devices in two directions. One is to higher speeds and densities, boosting functional integration and taking power savings to new levels. The other is to support the new next-generation capabilities and technologies such as Sonet/Ethernet convergence, GFP, and G.709 Forward Error Correction (FEC).

Higher Speeds

Just about everything is getting faster as metro line speeds home in on 10 Gbit/s. VT1.5 (1.5 Mbit/s) switch backplanes, for example, are now running at 2.5 Gbit/s and beginning to move to 5 Gbit/s. 10-Gbit/s network processors aimed at OC192 or heavy OC48 applications are starting to hit the market.

Vitesse Semiconductor, for example, plans to introduce its 10-Gbit/s IQ10G single-chip software-programmable processor family in mid-2002. Rival Applied Micro Circuits Corp. (AMCC) (Nasdaq: AMCC) announced in October 2001 the 10-Gbit/s nP7510 Network Processor as part of its existing family of 10-Gbit/s traffic managers, framers, PHYs, and switch fabrics (see PacketLight Intros Packet-based Transport and AMCC Adds to Chip Family).

Greater Integration and More Functions

Vendors are also cramming more and more functions onto their chips, so that chips are replacing complete cards and equipment racks. “Over the last decade a lot of companies made a big business out of building very big T1 switches – and now we have built that into a single chipset,” says Vitesse's Schmitt.

Vitesse argues that support for high-capacity grooming at both the VT1.5/TU-12 and STS1/STM1 levels is crucial at the metro edge because of the demand momentum for leased-line voice, T1s, and Frame Relay connections. Highly functional chipsets, such as Vitesse’s TimeStream family of Sonet/SDH switch fabric and line card devices introduced in mid-2001, can allow systems designers to offer more features to their carrier customers (see Vitesse Chops Chips). For example, having smaller VT switches distributed throughout the network obviates the need to backhaul traffic to a big centralized VT switch.

“In crossconnects you drive density. And we are trying to drive density with a functionality set that is superior to anything else that is out there. Our switching is completely nonblocking and has a very straightforward programming interface,” says Schmitt. “You can reprogram it very quickly. So, if there is a fiber cut, you can reprogram the switch in an instant, whereas with a TST [time-space-time] architecture there is an algorithm that you have to run your new configuration through, and that takes time. Every second that you are computing this new configuration is another second that bits are sliding on the floor.”

TimeStream also automatically detects failures and restores the network within a millisecond, according to Vitesse. The chip monitors the signal BER (bit error rate) and automatically performs a protection switch if a threshold is exceeded – and notifies the switch by sending messages in the Sonet signal. Users can then consider how the network needs to be reconfigured.

AMCC says that its Ohio S4806 chip, introduced in mid-2001, handles the functions of three separate chips. A channelized OC48 framer and ATM/POS processor, the Ohio processes an STS48/STM16 signal or four OC12/STM4 signals and ATM or IP data for up to 48 channels with a granularity down to the STS1/DS3 level. A line interface can be configured as a Sonet/SDH-compliant protection port or can be combined with a built-in digital crossconnect to add and drop TDM traffic.

Lower Power Consumption

Power consumption is tumbling, as single chips replace multichip configurations. PMC-Sierra’s ARROW-2xGE Gigabit-Ethernet/Sonet mapper replaces discrete SERDES, Gigabit Ethernet media access controllers, Sonet mappers, and FPGA devices to map two Gigabit Ethernet channels into an STS48/STM16 channelized payload. The company says power consumption is less than 2W/port, or less than 50 percent of that obtained using separate devices (see PMC-Sierra Launches Arrow).

But there are tradeoffs between functionality and power consumption, and the chip vendors are not necessarily aiming at the lowest possible power levels. “What we are seeing is that, with the level of integration on a linecard, where we have one chip that does virtually all the functions on a Sonet linecard, the power consumption there is not too much of a problem. It is much more challenging on the switch card as you are driving density to 160 and 320 Gbits/s and beyond. Power there is much more important,” says Schmitt.

Vitesse’s approach is to use an architecture that requires slightly more power than competing designs, but which allows straightforward programming. “You are talking of 24W instead of, say, 20W, but you’ve just eliminated the need for multiple software engineers. High-power devices require you to anticipate the thermal problems and provide customers with a heatsink solution that meets central office or outside plant applications.”

Smaller Transceivers

The quest for higher Sonet port densities, lower power consumption, and lower costs starts right at the bottom of the chain, with transceivers. Sonet transceivers are becoming cheaper, smaller, and more power efficient.

“The increasingly complex routing protocols place a huge demand on the equipment designers to maximize the availability of space inside their boxes for the higher levels of processing power (and memory) necessary to implement these functions. At the same time, the equipment designers are driven to minimize the space consumed by physical layer functions that add no value to their products,” says Bob Scharf, VP of marketing at transceiver vendor Stratos Lightwave Inc. (Nasdaq: STLW).

Stratos has introduced a new proprietary "RJ Format" for its optical transceivers, which it says reduces the card area required by 56 percent compared to multisource agreement Small Form Factor or Small Form Factor Pluggable transceivers. Further, because the RJ Format transceivers share the same equipment cabinet panel openings as the ubiquitous RJ-45 UTP connectors, system mechanical designers can reuse existing Ethernet or T1/E1/J1 rack and cabinet housings without having to redesign a new cabinet for each new optical system developed (see Stratos Launches Transceivers).

“This may sound like a rather pedestrian advantage, since it doesn't involve any radical new technology,” says Scharf. “However, cabinet approvals involve extremely expensive Telcordia, FCC, TÜV, and other safety-related test procedures. The cost and lead-time of these safety approvals is a continuous burden on the equipment suppliers. Thus, reusing existing equipment cabinets is a high priority in fast-time-to-market environments.”

Stratos has licensing agreements with vendors such as Agilent Technologies Inc. (NYSE: A) and JDS Uniphase Corp. (Nasdaq: JDSU; Toronto: JDU), and is setting up a multisource agreement committee. It is aiming the RJ Format at applications up to OC48 initially, mainly for the metro edge.

“We are positioning it as something that is really targeted at volume opportunities. That is where the bread-and-butter applications are right now. There is lots of talk about OC192, but it is still more of a research curiosity than a widespread protocol with a high-volume use anywhere,” says Scharf.

But activity is not confined to new transceiver formats, and many other transceiver vendors are pushing standard formats further. AMCC introduced its S19206 and S19210 transceivers in July 2001 for OC192 Sonet/SDH and 10-Gigabit Ethernet applications, and these can be used in 200-pin MSA or 300-pin MSA-based transponders/optical modules. The 0.13µm devices are designed to give high performance for edge rates, signal integrity, and input sensitivity, while consuming low power (see AMCC Delivers Transceivers).

More Flexible Transceivers

Chip vendors are beginning to build migration capabilities into their Sonet transceivers, and Intel Corp. (Nasdaq: INTC) highlights another twist to the Sonet migration issue through its TXN13303 transceiver, released in February 2002. This operates at 9.95-, 10.3-, and 10.7-Gbit/s by incorporating three customized jitter filters, and supports OC192 Sonet/SDH, 10-Gigabit Ethernet, and FEC-enabled optical systems, according to the speed setting. The idea is that systems designers can use one chip family to accommodate a move from Sonet-based to Ethernet-based MANs.

"This technology allows our customers to take advantage of the convergence of data rates at 10-Gbit/s in MANs," says Michael Ricci, vice president and general manager of Intel's Optical Products Group. "As Ethernet migrates into MANs, changing protocols will be as easy as flipping a switch."

Intel says that its aim is to encourage the adoption of next-generation technologies in the optical market by driving down costs and increasing integration to enable new capabilities. Adds Malcolm Hay, European marketing manager for Intel’s Optical Products Group: “We are delivering products at both the silicon and module level, which enables us to aggressively attack costs on two fronts. For the short/mid term our primary optical focus is at 10 Gbit/s, where we see the opportunity for Intel to bring the economics of mass-volume semiconductor manufacturing to the optical market.”

Sonet, Ethernet, and GFP

The biggest stakes are riding on the second branch of next-gen Sonet development – the new capabilities such as Sonet/Ethernet convergence via GFP and VC.

“If you want to be a player in Ethernet Over Sonet, you have got to be developing GFP chipsets, and you need to support a lot of functionality,” says Vitesse's Schmitt.

GFP chipsets are just beginning to come onto the market in some numbers. Agere Systems (NYSE: AGR), Agilent Technologies, PCM-Sierra, and TranSwitch have announced products over the last few months, and vendors such as Intel, Multilink Technology Corp. (Nasdaq: MLTC), and Vitesse are developing GFP chips. So expect to see a fair choice by the third or fourth quarter of 2002. PMC-Sierra claims to have been one of the first in the field.

“I think right now GFP and Virtual Concatenation are critical. We are getting a lot of traction on the ARROW-2xGE, as people want to roll out systems that can utilize Sonet as the transport mechanism, but be able to insert and extract Gigabit Ethernet streams into that,” says PMC-Sierra's Perna. “Previously, you’d have to use proprietary mapping technology to map only a single port of GigE into a single OC48 pipe. Now we can use Virtual Concatenation to not only map two ports of Gigabit Ethernet into a single OC48 pipe, we can virtually concatenate the remaining Sonet bandwidth together and make better use of that for multiple Sonet transport.” (See PMC Pushes Sonet Silicon.)

Chipsets are becoming available for Ethernet over Sonet that handle both the GFP and X.86 protocols (for example, see Agilent Boosts Ethernet-Over-Sonet).

Sonet and Digital Wrappers

Chip vendors are also pushing their Sonet designs to accommodate migration to future transport technologies. Vendors such as Intel, Multilink Technology, and Vitesse have processors offering combined Sonet and G.709 digital-wrapper functions that form the basis of the ITU-T Optical Transport Network (OTN), proposed as the basis of future wavelength-aware networks.

Says Dave Huff, VP of product marketing at Multilink, “On a physical layer side, most chipsets will work in either 10.7 Gbit/s for G.709 or 9.9 Gibit/s for Sonet. Once you have a Sonet network, if it’s just a DWDM point-to-point connection, it’s not really critical whether the link is managed via Sonet or digital-wrapper overhead – most people manage by Sonet at the moment because that was the technology that was available. However, for transparent transport of services, we are starting to see that the digital wrappers may be a more appropriate technology.”

An attraction of G.709 digital wrappers is their use of G.975 FEC to reduce the effects of transmission errors induced by noise, nonlinearities, and dispersion in the optical path. These are increasing problems in the U.S., where much fiber is old and of poor quality, by today's standards, and carriers are pushing metro core rings up to diameters of the order of 200km. FEC promises to increase transmission ranges and lower costs by reducing the need for repeaters and fiber upgrades.

“In general, in most of the G.709 chips that I have seen, the G.709 framing functionality comes pretty much for free with the FEC functionality. Today, most people are buying those chips for the FEC, not for the G.709,” says Huff.

G.709 can be used in a single-wavelength mode like Sonet, which it can transport transparently. A G.709 chip such as Vitesse’s VSC9271 also provides Sonet/SDH, Gigabit Ethernet, and Fibre Channel performance monitoring. Says Schmitt: “If you turn off the OTN the chip becomes a really good Sonet performance monitor, and maybe, in two or three years when you start converting that link to OTN, you flip a bit and suddenly you have an OTN-capable circuit pack.”

Next Page: Using Next-Generation Sonet

Next-generation Sonet is just part – although potentially a considerable part – of a metro solution. How, and to what extent, carriers will use next-gen Sonet will depend on many factors, such as carrier type, current network conditions, market segments targeted, financial conditions, and so on. Different flavors of next-gen Sonet have different capabilities, and carriers have to match these to their requirements and future plans.

Fortunately, carriers may have some leeway, as next-gen Sonet equipment is more flexible than classical Sonet. Vendors may, for example, offer alternative metro transport mechanisms for their MSPP products, giving carriers flexibility in their use of next-gen Sonet technologies.

Thus, the XDM multiservice optical network platform from Lightscape Networks Ltd. provides options for the underlying metro transport. “It can give a digital wrapper, or, if the customer prefers, Sonet/SDH framing,” says Ido Gur, VP of marketing. “Gigabit Ethernet cards are available in several options. One option is to put Gigabit Ethernet over a lambda directly. Another is to take the Gigabit Ethernet and map it into the Sonet/SDH metric, and then, of course, groom it with the Sonet/SDH bit rate. This Gigabit Ethernet has bandwidth control for selling partial Gigabit Ethernet.”

And many vendors clearly believe they have knockdown arguments for next-gen Sonet in ILEC (incumbent local exchange carrier) metro networks. These are already Sonet and carry huge amounts of TDM traffic – voice, Frame Relay, and private lines, for example – that is the cash cow that has helped them fend off the threat from the CLECs (competitive local exchange carriers) and other data/Ethernet-only carriers. The ILECs’ great problem is how to enter the Ethernet services market profitably, without disrupting the TDM side. Next-generation Sonet is proffered as the solution.

Says Lucent's Goode: “The approach that a product like Metropolis DMX takes, where you have basically what amounts to a next-generation Sonet multiplexer to provision your voice and private-line services, is that all the Ethernet technology and components are contained on the card. Ethernet then becomes a pay-as-you-grow service upgrade. You plug in a 10/100-Ethernet card, say, and you can offer Ethernet services over your existing Ethernet infrastructure. And ILECs have found that to be a very cost-effective way to add support for data services to their networks.”

The basic use of next-generation Sonet is to provision and control bandwidths effectively for a range of service types. Typical are:

Paradoxically, a useful way of viewing next-generation Sonet may be as a kind of carrier’s Ethernet. Strange, since Ethernet in its Gigabit manifestations has caused a complete rethinking of metro architectures and Sonet’s role; but both next-gen Sonet and Gigabit Ethernet are being evolved on widely implemented, understood, and well liked technologies. Both are responses to technology changes, and both have acquired additional capabilities far beyond anything their original designers intended.

Such capabilities mean that Sonet has a lot of life in it from the mix-and-match point of view. Just using current standards, Sonet can currently handle, for example, ATM, IP, Ethernet, Frame Relay, and Ethernet-over-Sonet with Virtual Concatenation and GFP – and with native-mode protocols such as Fibre Channel and Escon via GFP’s transparent mode.

"Sonet/SDH is going to continue to be the premier transport protocol for at least the next five years,” says Vitesse's Schmitt.Critically, the more capable next-gen Sonet has the potential to act as a type of metro service convergence layer through a combination of protocol transparency and highly granular switching and bandwidth management.

“If you look at history as an indicator, ATM as a Layer 2 technology provided interesting capabilities, like traffic engineering and bandwidth management, and received a lot of hype which suggested it could do much more. At one point it looked like it was going to take over the world, and it was expected to be the service convergence layer," says ONI's Duffy.

"ATM unfortunately lacked transparency, which is why it really missed the mark in terms of providing convergence. In other words, ATM could not support other protocols in their native formats, which prevented it from scaling well. And MPLS went through a similar cycle in terms of being suggested as a convergence layer, and now we are watching Ethernet repeat history, with RPR going through yet the same cycle once again.

“I think we should learn from history and try to use a technology for convergence that is transparent and adds simplicity to the network. A combination of optical technologies, as opposed to any Layer 2 protocol, will provide that capability for carriers.”

Next Page: Systems Developments

Pretty well everyone in the Sonet business is working on next-generation Sonet or closely related areas. The following table displays some of the principal vendors and their products, to give a flavor of where Sonet development is going. Basically, it’s about boosting Sonet’s traditional capabilities and making better use of single-wavelength capacity for service creation and traffic management in an increasingly packet-oriented environment. (We've stopped short of getting too involved in DWDM, which will be the subject of a future report.)

For each of the vendors covered, the table takes a sample metro product (or products) and lists some of its major characteristics under several broad headings:

Type: How the vendor describes the product

Function: Many of the products are highly integrated, combing ADM, digital crossconnect, data switching, and DWDM terminal functions

Next-generation Sonet base: Vendors vary considerably in their use of the various next-gen Sonet technologies. All use some form of concatenation, but this may be contiguous or virtual; and the mapping/multiplexing scheme may be proprietary. Of the new standard schemes, GFP looks to be gaining popularity (for example, Astral, Lucent, ONI); X.86 is more limited (Atoga), as is RPR (Nortel, Siemens). But many vendors do not specify a standard.

Traffic/service interfaces: The main interfaces provided for network-side Sonet and client services. These are influenced by market segment, metro application, and the underlying next-gen technology, so there is a lot of variation in the details. A basic set that is common to many of the products is DS1, E1, DS3, E3, OC3/STM1, OC12/STM4, OC24/STM8, OC48/STM16, OC192/STM64, Gigabit Ethernet – in other words, the obvious key TDM and OC interfaces, with Gigabit Ethernet as the focus of the packet-data offering.

Vendors appear to split on how they build on this basic set. Some (such as Astral, Atoga, Lucent) concentrate solely on the Ethernet side by including 10/100 Ethernet and 10-Gigabit Ethernet; others (such as Alidian, ONI) widen the range to include newer protocols such as Fibre Channel and/or legacy protocols. Broadly, the emphasis of most of the sample is on supporting the various forms of Ethernet.

Client service capabilities: Given the selected vendors’ emphasis on Ethernet as the main packet-data interface, Ethernet capabilities dominate this category. Point-to-point LAN transport, multipoint LAN interconnect and extension, and VLANs are the most common basic set. But the vendors differ in the further capabilities they emphasize – per-customer and per-application SLA; IP flow classification, policing, and metering; multiple classes of service and COS-based queuing are all offered by Atoga, for example.

Scaleability: How client data services and/or systems can be scaled, especially crossconnection, grooming, and Ethernet granularity. 50-Mbit/s (STS1) crossconnect/network granularity is standard across all the relevant products; and VT1.5 granularity is offered by a significant proportion (for example, Nortel, Turin). Ethernet client-service granularity varies quite widely – 50- or 100-Mbit/s increments to 1-Gbit/s are common; but some vendors (for example, Nortel, Tellabs) have 1- or 1.6-Mbit/s increments over this range.

Network protection: All vendors support at least one of the standard Sonet ring automatic protection schemes – usually USPR – and many support both USPR and BLSR. Many offer further options (such as 1+1 linear), and some offer Ethernet spanning tree (for example, Cisco, Lucent).

System reliability: Typically carrier-grade, > 99.999%. (This is Sonet, after all.)

Physical configuration: Basic chassis size, cards per chassis, chassis per bay, and ports per card. Physically, there is a common approach of using a core chassis, shelf, or system into which interface and functional cards are slotted; generally, interface cards can be mixed and matched by type until the capacity limit of the chassis backplane is reached. Chassis sizes vary considerably, with two, three, or four fitting into a standard 7-foot bay (19 or 23 inches), together with the necessary power-supply and ancillary units. The number of free client-service card slots per chassis varies between about six and 12, but the number of ports per card tends to be less variable – many of the systems support two or four Gigabit Ethernet ports per card, for example. The maximum number of ports of a given type per rack varies as well. Many of the systems are in the range of 40- to 72-Gigabit Ethernet ports per rack.

Power consumption: Per-full-rack power consumption depends on specific card and chassis configurations. Based on vendors’ typical figures, the products typically lie in the approximate range of 1.5 to 3.5 kW.

Standards: Major environmental and other standards supported. All the products are aimed at the established Sonet market, so they almost invariably are specified to such standards as NEBS (GR-63-CORE, GR-1089-CORE Level 3), UL 1950, and so on.

Vendor approach/notes: Our take on each vendor's design approach and emphasis; other relevant points



Dynamic Table: Selected Systems Developments

Select fields:
Show All Fields
Vendor ProductTypeFunctionNext-Generation Sonet base Traffic/service interfaces Client service capabilitiesScaleability Network protectionSystem reliability Physical configurationPower consumptionStandardsVendor approach/notes

It’s impossible in the space available to include all vendors’ complete Sonet metro ranges, as these usually span several several functional categories – premises access, access nodes, aggregation nodes, core nodes, and so on. Also, quite a lot is simply offered as classical Sonet – improved and higher density, to be sure, but not really relevant to the spirit of this report.

Applying various next-generation Sonet technologies across the whole range of metro functional categories (not to mention DWDM approaches) produces a lot of possible product approaches. Combine this with a market segmented by carrier type – ILEC, CLEC, data-only CLEC, and so on – and different approaches to metro architectures and functional integration, and the result is that the next-generation Sonet product universe is populated by too many apples and oranges for a simple comparison of a few basic characteristics to be very meaningful.

So the table should not be read as a straight comparison of vendors’ products. But it does highlight some of the different outcomes that next-generation Sonet is producing in product terms – and what you can expect to find on the market now and over the coming months.