1999's 10 Hottest Technologies
Service Mediation
Metro DWDM
OC-768 SONET
HDSL2
Unified Messaging
Policy-Based Management
MPLS
Smart Antennas
Optical Switching
Wireless Location
10 hottest technologies of 2000:
1. Soliton
2. Passive optical networks
3. Voice over DSL
4. Time Division Duplexing
5. Tunable Lasers
6. Optical Domain Service Interconnect (ODSI)
7. Synchronization
8. IP VPNs
9. Gigabit Ethernet
10. Very-short-range optics
Soliton: Barging Ahead
The concept is more than 150 years old, but it could revolutionize optical networks. Although not without controversy, some vendors are looking to solitons to build the next generation of dense wavelength division multiplexers (DWDM) that can operate over longer distances and provide higher-capacity channels. Solitons are pulses of light that keep their shape even after colliding with one another. This phenomenon was first observed by a Scotsman, John Scott Russell, as he rode alongside a barge in a canal in Edinburgh. As the barge came to halt, he noticed that a large solitary wave, after colliding with other waves, continued for several miles without losing its shape.
Although work with solitons began a quarter-century ago at Bell Labs, advances in DWDM technology have overshadowed this research. Now some companies such as French start-up Algety Telecom are convinced soliton DWDM systems will solve the problem of chromatic dispersion, or the broadening of a pulse of light, in DWDM networks. Algety, which has its roots in CNET--France Telecom’s R&D unit that is credited with being the first to transmit 1 Tbps over 1000 km using soliton technology--plans to have a beta version of a soliton-based DWDM system before the end of the year. The company is particularly interested in the U.S. market and will soon be establishing an office in the States. According to Jereme Faul, Algety’s vice president of software strategy and development, while there might be some debate over whether solitons are necessary for 10-Gbps wavelengths, 40-Gbps and higher wavelengths demand the use of solitons.
However, officials at Cisco Optical Networking Group (formerly Pirelli Optical Systems), have rejected solitons for high-rate systems, although MCI conducted a successful trial in 1998 using Pirelli’s soliton-based DWDM systems that transmitted four data streams at 10 Gbps each. Stuart Woods, North American product marketing manager for the Cisco group, noted that soliton uses the return-to-zero (RZ) modulation format, whereas network devices such as routers, ATM switches and SONET equipment are based on the nonreturn-to-zero (NRZ) scheme. “Once you go down the path of RZ, you’re automatically closing yourself to being interoperable,” he said. According to Woods, neither distance nor faster channel speeds require the use of RZ. Moreover, a move to 25-GHz channel spacing cannot be done with solitons because a soliton has wide pulse. According to Woods, what exist today are not true solitons but dispersion-managed solitons that require properties in the fiber to tweak the pulse so it holds its shape. “Dispersion-managed soliton is what we used in the trial [at MCI],” he said.
Rick Barry, chief technology officer for Sycamore, takes issue with such criticism. Barry said that there is “overwhelming agreement” that dispersion-managed soliton is the choice technology for next-gen, ultralong-haul, terrestrial systems. Addressing the objection to the RZ format, Barry said: “Most routers use NRZ because NRZ is simpler and routers were not designed or required to transmit ultralong-haul distances. The interface between the routers and WDM system will convert the NRZ format to the RZ format. This [approach] separates the design of the routers from the design of the WDM system.” Algety received a vote of confidence on the technology in March when a second round of venture capital funding raised $31 million.
PON: Sharing the Light
The potential of the Internet remains tied to solving the issue of the last-mile bottleneck. While much attention is being given such broadband access technologies as DSL, cable modem and wireless, often overlooked is the passive optical network (PON) despite indications of a fast growing fiber access market. Communications Information Researchers (CIR) estimates that the combined fiber-to-the-curb/fiber-to-the-building market in the United States will jump from about $800 million this year to approximately $1.8 billion in 2004. PON technology, which only came to public attention last year, provides a straightforward way for facilities-based service providers to support customers’ multiservice needs.
Put simply, a PON consists of an optical line terminal (OLT), which resides at the central office (CO) or a point of presence (POP), and an optical network terminal (ONT) located on or near the subscriber. The ONT provides the various service interfaces, such as Ethernet or DS1, to users. In the downstream path, a passive optical coupler, which can be located anywhere between the OLT and ONT, slices the light to serve as many as 32 end points and, in the upstream direction, combines the light. Some couplers might split the optical power evenly, others might do a directional split where, say, 90 percent of the power goes down one path and 10 percent down a second.
There is enough interest in the PON that a number of carriers, including SBC, US West, BellSouth and Bell Canada, are participating in the Full Services Access Network (FSAN) consortium to create a comprehensive PON standard. This group decided to use ATM as the transport technology in the PON and this ATM PON format is now the ITU G.983 standard. But while this standard specifies key features such as operation over single/dual fiber operation, a range of 20 km and certain data rates (currently the maximum is 622-Mbps downstream/155-Mbps upstream), implementations by vendors such as Quantum Bridge, Terawave Communications and Optical Solutions Inc. may go beyond such basics. For instance, the Quantum Bridge solution, which is in trials by cable giant Comcast, is optimized for IP although the company also supports ATM for voice traffic. Similarly, Optical Solutions’ FiberPath offering can support six phone lines, 135 CATV channels and scalable high-speed data on the fiber pair running between the splice point and the subscriber.
The beauty of a PON is that there are no active components between the CO and the customer’s location and hence there is no requirement for power or maintenance. Today a PON network splits the light to provide services to multiple customers. But as DWDM technology moves beyond the network core and into the metro, a PON could prove equally adept at delivering an entire wavelength to a customer. In the near term, a PON could also complement the rollout of DSL services. Tom Nolle, president of CIMI Corp., noted that BellSouth is interested in the ATM PON to feed a fiber remote from where DSL or customer fiber might be used to complete the connection to the user. “It’s very likely that some form of PON will grab a big chunk of the U.S. access market in the long term. In the near term, it’s not clear how it fits in the regulatory framework for the RBOCs. If it’s used to feed a fiber remote, I think it would be a killer strategy even today,” Nolle said.
VoDSL: Extending Copper’s Economics
The CLECs’ focus on digital subscriber line (DSL) has centered around the technology’s ability to deliver high-speed Internet access over copper. But VoDSL is proving CLECs need more than just data to survive in the small to medium-size business market. While this market is nascent, data CLECs may find themselves struggling to compete without a key characteristic like VoDSL, which enables them to expand their customer reach with a double-fisted service on one line. Mpower Communications, one of the early adopters, argues that the move to VoDSL is crucial: “The economics are great,” said John Boersma, senior vice president of engineering and network operations for Mpower. “We think that a VoDSL package is extremely important because we expect to see continued downward pricing pressure on pure data DSL.”
A host of gateway vendors and carriers have been racing forward with their own flavors of VoDSL. Gateway vendors Accelerated Networks, CopperCom and Jetstream all utilize ATM in their solutions; Tollbridge has opted for an all IP-based gateway. Some of the early adopters include Gateway Communications, Mpower, Network Plus, Picus Communications and Rio; other large CLECs such as Rhythms NetConnections and Covad are in the process of testing the service.
A basic VoDSL system consists of three elements: an IAD, DSL access multiplexer (DSLAM) and a voice gateway. DSL connection to the customer makes use of a packet protocol (generally ATM) to support both voice and data. The DSLAM functions as a packet concentrator, delivering traffic from multiple customers over a high-speed uplink to a metropolitan or a regional packet network. Voice services are delivered to DSL customers through the voice gateway, which connects to the PSTN via a GR.303 interface.
Despite VoDSL’s potential success, analysts are quick to point out the reliability issues vendors need to overcome to enable large-scale deployment. “For an enterprise considering VoDSL, there are several major advantages: lower monthly bills, a single integrated voice and data network and a wider array of local voice operators from which to choose,” said Brett Sheppard, a senior analyst with TeleChoice. “Downsides include having to rely on battery backup of voice lines, or cell phones, in the event of a power failure and possible problems in services such as caller ID that are highly sensitive to latency over backhaul circuits. For many small to midsize businesses, lower monthly calling bills and an integrated network will offer convincing arguments for an enterprise CEO or CTO to sign off on VoDSL.”
According to the Yankee Group, VoDSL will not make a major penetration until 2001. Yankee predicts that 5 percent of projected DSL business subscribers will also make the transition from standard voice services to VoDSL. And, Yankee postulates, if these business subscribers average $400 per month for basic voice and data services and generate $320 per month in value-added service revenues, this equates to potential revenues of roughly $475 million generated in 2001, expanding to just under $3 billion in 2003.
Time Division Duplexing: Maximizing Spectrum
Like the CDMA-TDMA-GSM debate in mobile wireless networks, a similar battle is brewing in fixed wireless: frequency division duplexing (FDD) vs. time division duplexing (TDD). But unlike the mobile world, action in the local multipoint distribution service (LMDS) spectrum, with the exception of two or three notable carriers, has been a case of all talk and less-than-widespread deployment. Air interface standards, currently being worked on in the IEEE’s 802.16 committee to deal with issues such as interference and interoperability, could help change that, propelling broadband wireless to a $7.4 billion market for services by 2003, according to the Strategis Group.
The deployment of TDD systems could be a big boost to the market as well. While the major broadband wireless deployments from Winstar and Teligent have been tried-and-true FDD, these and other carriers, such as NextLink, are taking a close look at TDD for its spectral efficiency capabilities. TDD, which is being backed by start-ups Ensemble Communications and WavTrace, was designed with data traffic--asymmetrical and “bursty” by nature--in mind. TDD uses a single channel for simultaneous, two-way communications. Separation between the transmit and receive functions occurs in the time domain rather than by frequency (as in FDD). With TDD the radio’s uplink and downlink operate on the same frequency, but at different times at a fixed interval and the amount of bandwidth allocated to each direction is flexible. This circumvents the need for a large guard band to isolate the upstream and downstream traffic, increasing the efficiency of the spectrum.
The LMDS spectrum is quite large, so efficiency is not currently a factor: If service providers have both A and B band channels in a market that’s a total of 1150 MHz of spectrum, equivalent to about 200 video channels. “Service providers are really just starting to sign on customers, so it’s not an issue just yet,” said Andy Fuertes, senior analyst with Allied Business Intelligence. “But going forward, as broadband wireless catches on, then carriers may be thinking a lot more about spectral efficiency.”
Just as analysts believe there’s enough room for broadband access and mobile wireless via several technologies, so do they think there’s room for both FDD and TDD. Whether TDD has a compelling enough story to dominate the market is up in the air, but the interest being paid to the technology is positive. NextLink, the largest LMDS spectrum holder in the country, has chosen WavTrace for its Los Angeles trial along with three other vendors. The start-up also has a strong relationship with Harris Corp., which owns nearly 20 percent of the company. Ensemble has strategic marketing alliances with Digital Microwave and ADC.
Tunable Lasers: The Next Wave
Tunable lasers may become the crown jewel in the all-optical network’s treasure chest. While multichannel DWDM systems have pushed fiber’s capacity to transmit multiple wavelengths, the necessary cost of replacing or keeping spare fixed lasers is becoming unattractive. With a tunable laser, carriers will only need to tune to the appropriate DWDM wavelength and insert a card instead of using a transmitter for each wavelength. In a 32-channel DWDM system, for example, a carrier could cut the number of lasers from 32 to four separate lasers that can tune more than eight wavelengths each.
Beyond sparing capabilities in the long-haul market, tunable lasers can be used in add/drop multiplexers in conjunction with tunable filters. Vendors predict that tunable lasers that perform nanosecond switching from wavelength to wavelength will enable all-optical network switches and wavelength routing. They are also beginning to realize the potential of the metro market and its specific needs.
“The tunable laser market is very real,” said Robert Plastow, chief technology officer of Altitun AB. “It’s still segmented a little bit in that if you ask where will the first tunable lasers be used in networks, it’s going to still be sparing in long-haul. But wavelength routing and the metro stuff are moving tremendously fast. That’s where a lot of our backlog is coming from.”
Several traditional and upstart vendors, including Agility Telecom, Agilent Technologies, Altitun, Bandwidth9, CoreTek (recently acquired by Nortel), Fujitsu, Lucent and NTT have been working to develop a cost-effective, tunable laser solution and are beginning to make good on their promises. Like any emerging technology, vendors are already beginning to debate the best tuning method. Currently, there are four: distributed feedback lasers (DFB), distributed bragg reflector (DBR), MEMS (microelectromechanical switches) and temperature variation. DFB lasers supply necessary power for long-haul applications. DBR lasers keep the established processing and packaging technologies, but add nanosecond wavelength switching and a broad tuning range. Other vendors are abandoning telecom laser technology in favor of mechanically moving parts, either as conventional external-cavity lasers or as smaller MEMS and vertical cavity designs. Vendors like CoreTek have opted to take a fixed-wavelength laser design and vary the temperature of the laser chip.
Analysts agree that the potential for tunable lasers is real, but are quick to point out that vendors need to develop solutions that are at least the same price or lower than that of fixed-wavelength lasers. “There is a lot of interest in tunable lasers, but the main issue holding them back is the need to get the costs down,” said Lynn Hutcheson, director of optical components at analyst group RHK. “The fixed laser market alone for 1999-2000 is huge, with 450,000 units expected to be sold. Once the tunable laser market takes off, it may make up at least 15 percent to 20 percent of the total laser market. So far the only company to successfully demonstrate the technology is Altitun, which completed a field trial with Telenor. Many of the other emerging players need to meet Bellcore standards before they release any products.”
ODSI: Closing the Electro/ Optical Gap
While DWDM will continue to push fiber’s capacity to unprecedented limits, it does not provide an intelligent method for optical components to communicate with their electronic counterparts. Enter the ODSI (Optical Domain Service Interconnect) initiative. Pioneered by optical switch start-up Sycamore Networks, ODSI was formed in conjunction with a group of 50 service providers and optical switch vendors. Its primary goal is to define a practical framework for interoperability between electrical devices such as IP routers and ATM switches to make bandwidth requests directly to an optical network via optical switches. Initially, the consortium placed its energy on the development of an optical user network interface (OUNI), which is based on extensions to multiprotocol label switching (MPLS) standards.
By bringing immediate momentum to the issue of interoperability at the electro-optical boundary, service providers such as Williams Communications, which is an early supporter of ODSI, will be able to more effectively architect their multivendor networks to support the demand for more bandwidth and high-speed services. “ODSI will enable us to automatically provision bandwidth for our IP network,” said Andy Wright, chief technologist for optical networking at Williams. “It also has applications that go well beyond IP and for other types of provisioning, such as interoperability between transport networks and being able to provision end-to-end circuits across different mesh networks. One of our big initiatives is to build a meshed network with quality of service (QoS). Right now, this will be a one-vendor network; we will not be able to do multiple vendors because it is really a closed network architecture.”
Analysts said the gigabit/terabit switch router vendors participating in ODSI are headed in the right direction to drive efficiencies in the optical access layer. Although these companies have laid low for a while, they are on their way to reaching market penetration. “I think the terabit/gigabit switch router guys are the ones to watch,” said Chris Nicoll, director for carrier infrastructure at Current Analysis. “Comments from service providers that have talked to them about architectures, scaling, features and functionality have been very positive. [Terabit/gigabit switch router vendors] had a black eye, but the eye is healing and they are starting to hit their stride.”
If ODSI is successful it could be implemented by the major manufacturers of core IP routers, ATM switches and optical switch vendors. ODSI interoperability trials should begin in the fall.
Synchronization: It’s About Time
With all the attention given voice over IP (VoIP), it’s a wonder there is so little discussion about synchronization. Synchronization? Yes, because while IP may prove adaptable to delivering the QoS necessary to handle voice and video traffic, the facts remain that the existing circuit-switched network is not going away any time soon and that the IP network and the PSTN will have to coexist. Besides VoIP, legacy applications such as fax and analog modem will require synchronization in a hybrid IP-PSTN environment.
The concept of synchronization, or network timing, is more easily understood by analogy: Without traffic lights, downtown traffic would come to a standstill. Although adding lights would alleviate the situation, things would be better if the lights were synchronized. In short, synchronization ensures that information is received at similar rates as it was transmitted. If this does not happen and the network is unable to buffer the information, the result could be degradation of some services (compressed video and encrypted voice) or worse, loss of information (Group III fax).
Synchronization is accomplished by deploying clocks in the network. In the old days, a single clock in a vault in Missouri provided the timing for the entire North American telephone network. Now there are several types of clocks, including stratum clocks, which are based on different performance levels. There are basically four strata for performance, with Stratum 1 clocks sitting at the top of the heap. Synchronization of the PSTN is generally accomplished via primary reference source (PRS) clocks, which generally derive their timing from the global positioning system, and Stratum 2/3 building integrated timing supply (BITS) clocks--referred to as synchronization supply units outside of the United States--that typically receive their timing from the PRS or the network itself. “When we went from analog to digital, synchronization became critical; then as data rates went up, it became more critical; and when SONET got deployed it became even more critical,” said Murli Thirumale, Symmetricom’s vice president of marketing and new business development. “Now with packet networks coming in, it’s essential that the underlying clock mechanism be of high quality.”
Fortunately, with the pricing of synchronization technology coming down, the timing network is becoming less hierarchical. “Prices today are low enough that you could put a Stratum 1 clock in every central office, perhaps using a Stratum 3 clock as a backup,” Thirumale said. Indeed, what used to be a chassis-based product has been reduced to a set of smart chips by Datum-Austin. These self-configuring chips can be embedded into equipment targeted at service providers. “I would guarantee that a symposium on synchronization could be given by any of the carriers today,” said Jack Rice, president of Datum-Austin. “But guess what, the customers that the data equipment vendors are now targeting are the carriers. If these vendors imbedded a clock in that buffered data product, they’d no longer have to worry about interfacing with the nonbuffered, circuit-switched world.”
But will these clocks only continue to have a play in the transition from circuit- to packet-switched networks or is there life for them beyond that? Some say synchronization will be important as long as there are applications that require constant bit rate service. David Schwartz, senior analyst at Dataquest, believes it will increase as network elements proliferate and features or services are enabled by feature servers rather than being embedded within large network elements. “As this distributed architecture evolves, it will be extremely important to synchronize these elements in order to guarantee QoS,” Schwartz said.
IP VPNs: Virtual Routing Takes the Stage
With the advent of virtual routing, virtual private networks (VPNs) will just about exit the customer premises and reinvent themselves as network-based services with all the trimmings IP can deliver. While traditional VPNs consist of packet transport and customer premises equipment that perform encryption, firewall and other security services, virtual routers sit at the edge of the carrier’s network and serve as the building block to offer scalable IP VPN services from the carrier cloud. They don’t demand large conversions from customers because they can use existing addressing schemes. Nor do they require costly provisioning of each end-user site and complex CPE. Providers will be able to offer boilerplate firewall, encryption and intrusion-detection services as well as self-serve portals and other application-level provisioning and monitoring to multiple customers.
According to Ron Westfall at Current Analysis, only three vendors can provide network-based IP VPNs from the carrier cloud: CoSine, Shasta (acquired by Nortel) and Spring Tide, because they have built platforms to support wide-scale virtual routing via multiprocessor architectures and advanced OS software. Traditional routers still lack the ability to segment and layer services to discrete subscriber networks, he said. However, don’t dismiss Cisco and Redback, which have hinted at big plans. CoSine’s Mark Showalter, vice president of marketing and product management, is aware that MPLS is being positioned to compete with IP VPNs but he doesn’t consider this technology in the same league. He thinks that by creating what is essentially a frame relay PVC, MPLS is just reiterating private line services. He said it lacks the dimension to offer secured access to the Internet and extranets to different sites.
The proving ground rests with service providers: Can they deploy network-based IP VPN services with service level agreements (SLAs) to back them up? CIMI Corp. has estimated the IP VPN market to reach $40 billion worldwide over the next 10 years. Qwest is confident it has the savvy and is building a multimillion dollar IP platform with CoSine equipment that will be completed this year. Application-layer solutions for IP VPNs are already available from vendors such as Abatis Systems, which in March announced a self-serve solution for on-demand provisioning, bandwidth allocation and extranet creation.
10-Gigabit Ethernet: Riding the Cost Curve
Tried-and-true Ethernet will make its way to the WAN later this year in the form of 10-Gigabit Ethernet (10 GigE). The IEEE 802.3ae task force plans to have a draft standard based on the well-recognized protocol out in September that is compatible with WAN and LAN infrastructures, eliminating time-consuming protocol conversions. Customer trials will be later this year and prestandards commercial products are expected in Q1 2001 from vendors including Foundry, Nortel, 3Com and Extreme Networks, which are four of the seven founders of the 10-Gigabit Ethernet Alliance formed in January.
10 GigE is an OC-192, short reach, 40-km technology that plugs directly into a service provider’s optical network via a switch/router interface, making it a good fit for intra- and inter-POP connectivity, server farms or for extending a POP footprint. Service providers will be able to offer seamless, scalable end-to-end solutions for high-speed access from the WAN to the LAN and, for the first time, a leased Ethernet solution.
The real reason 10 GigE will win is cost: It has a lower price tag relative to ATM and SONET. Dell’Oro Group predicts that price per port for 10 GigE is 8.3 times the price of a single gigabit port, which means buying one 10 GigE port will save 17 percent over the price of 10 single gigabit ports. The 10 GigE market will reach $1 billion by 2004, according to the research group. IDC’s Esmeralda Silva-Swartz cautions, however, that 10 GigE lacks SONET’s link-management capabilities that permit users to track link failures. There are some proposals to use a digital wrapper to encapsulate Ethernet frames to provide this capability, but it would have to be done with nominal incremental cost and complexity, according to Silva-Swartz. Packet over SONET (POS) may also have some advantages over the long haul, but Nortel is banking on Ethernet’s ability to handle bursty traffic and mesh networks while POS is limited to point-to-point transport.
In fact, Nortel believes 10 GigE is the death of distance and will define the common building point for networks. “We see 10 GigE as the cheapest way to integrate the LAN, MAN and WAN,” said Rod Wilson, director of Nortel’s 10 GigE project. “The relevance of how one fences off these arbitrary definitions will no longer apply. It will be the network. And because of Ethernet as the one common protocol, the only issue will be scalability.”
Very-Short-Range Optics: Shorter is Better
Service providers are currently paying for SONET transport designed to reach up to 500 km when, in some cases, they really only need to cover 200 to 300 meters. Very-short-range optics, a SONET-framed interface that relies on parallel optic technology and will serve as a replacement to expensive serial interconnects, will help service providers cost-effectively address customers’ demands for intra-POP transport beginning at OC-192. The technology may displace traditional serial single-laser interfaces with an array of 12 850-nanometer vertical cavity surface emitting lasers (VCSEL). The laser array is fabricated onto a single chip, which amounts to the same cost as packaging one single-wavelength laser; service providers pay one-twelfth the cost of packaging and receive 12 times the capacity. From this converter chip, the signal is mapped over 12 1.25-Gbps fiber links, recombined and transmitted out at 10 Gbps. As a result of transmitting over multiple lines at a lower bit rate per line, the technology requires lower-cost optics and silicon, making it an ideal transport for router-to-router, router-to-DWDM terminal, or router-to-optical cross-connect (OXC) distances.
Cisco and Ciena are working on VSR optical solutions and have announced interoperability. Cisco views this first generation of VSR optics as a stepping stone although a standard short-range SONET serial interface may reach competitive price points within three years. Andy McCormick of the Aberdeen Group expects OXC vendors, including Xros (bought by Nortel), BrightLink (formerly Corvia), Sycamore and Tellium, to add VSR optics as a natural extension to their offerings. “The demand for VSR optics is growing because ASPs need to interconnect to carriers’ networks in data centers and new types of carriers, such as Yipes, are focusing on Gigabit Ethernet IP-based services,” he said. The Optical Internetworking Forum (OIF) will be voting on a draft standard this month. McCormick predicts VSR optical products will not be commercially available until Q2 2001. |
