Ten Hottest Technologies
These Picks Sizzle, Sparkle, & Shine
Cover Story May 2001
The Telecommunications® Staff
telecommagazine.com
--------------------------------------------------------------------------------
One of the hardest parts about putting together an annual list of the 10 Hottest Technologies is peering into the crystal ball and looking two years down the road at what the telecom industry will look like.
--------------------------------------------------------------------------------
In many ways it's like a service provider's Christmas wish list: We have bandwidth. We want more bandwidth and we want it in places that have become huge bottlenecks--the metro and access networks. We want more cost-effective solutions to move more bits over longer distances. We want a more effective way to groom and shape this ever-growing amount of traffic. Oh yeah, and we want a way to better provide enhanced services over new and incumbent networks and really start making some money.
Our choices for hot technologies are not a quick fix. They will take a year or two to hit the mainstream and really start making a difference. We have again asked some of the industry's leading analysts to validate our choices and their responses told us we were dead-on. Here, in no particular order, are Telecommunications' picks as the 10 hottest technologies of 2001:
UltraDWDM
Advanced FEC
VDSL
Free-space optics
Resilient Packet Ring
H.248/Megaco
Directory-based management
Adaptive modulation
Raman amplification
SCTP
1
UltraDWDM:
100s of Points of Light
Two years ago Telecommuni- cations® magazine accurately picked OC-768 as a hot technology. If 40-Gbps fat pipes are good, it should follow that more pipes are even better. Indeed, DWDM is about to morph into, to coin a term, UltraDWDM. Whereas vendors today typically support anywhere from 80 to 96 channels in the popular C band (and they can double those densities by having an equal number in the L band), they now are seeking to offer literally hundreds of channels within each of these bands by going to tighter spacings. For instance, Ciena last fall announced that it had successfully tested 25- and 12.5-GHz spacing. While Ciena uses internally developed fiber Bragg gratings (FBGs) for 50- and 100-GHz channel separation, 25-GHz FBGs may soon be available from multiple optical component vendors; Southampton Photonics claims that in February it became the first company to commercially ship 25-GHz FBGs. With 25-GHz spacing, vendors will be able to provide about 160 channels in the C band and twice that number by employing 12.5-GHz spacing. In fact, there's even some talk about 6.25-GHz spacing. And, of course, these close channel spacings can also be extended to the L and S bands to further increase a system's future capacity.
Such a channel bonanza, however, may come at the cost of sacrificing channel capacity. When data is encoded on a particular frequency, a band of additional frequencies are necessary around that "center" frequency to transmit the data. Furthermore, because the band's size is about equal to the bit rate of the signal that is to be carried, a 40-Gbps data stream would require about 40 GHz of spectrum. Consequently, some vendors say 40-Gbps transmission is incompatible with 25-GHz channel spacing. Moreover, the industry has used until recently a design rule saying that channels should be placed two-and-a-half times apart depending on the bit rate. So 10-Gbps channels would have a 25-GHz separation (although currently 50-GHz spacing is common) and 40-Gbps channels a 100-GHz separation. Improved components are expected to enable closer channel spacing and, consequently, more efficient spectrum use.
Not all vendors, however, are captivated by the idea of packing channels tighter and tighter. Start-up PhotonEx, for one, believes 40-Gbps channels are what carriers want for long- and ultralong-haul applications. On the other hand, Ciena (which has 10-Gbps channels using 12.5-GHz spacing in the lab), Sycamore and others believe there is a market for DWDM systems that can provide many channels economically, particularly in the C band where the technology is most mature. In the end, there may be no argument at all if it turns out there is not only a market for fat pipes, but also for lots of smaller pipes.
2
Advanced FEC:
Getting It Right
Squeezing more distance from the optical network has fast become the industry's holy grail, and FEC (forward error correction) is a key technology for enabling it. Hardly a new concept, FEC has been used for years to keep BER in check. The challenge for FEC in optical networks is providing a BER of 10-15 when transmission speeds are measured in gigabits/sec.
The FEC standard that's been around for years is the ITU-T G.975 specification, which uses the Reed-Solomon algorithm. Implementing G.975 in an optical network requires adding about a 7-percent overhead to the optical signal. Thus with G.975, a 2.5-Gbps transmission would be boosted to 2.7 Gbps and a 10-Gbps signal would be boosted to 10.7 Gbps. Now, however, OEM vendors and semiconductor chip suppliers are working to provide a new and improved FEC. There is no official name for this nonstandard FEC. Advanced FEC works, although some vendors more appropriately call it advanced code FEC or high gain FEC because, in contrast to G.975 which provides a coding gain of 6db, Advanced FEC provides approximately a 9db coding gain. The challenge is in developing a FEC that provides more coding gain but uses less overhead. For example, Ciena claims that the Advanced FEC it's developing will only add a 7.1 percent overhead. While some vendors are choosing the internal development route, others turn to chip suppliers such as Intel or Vitesse Semiconductor. Vitesse, which has a FEC and an Advanced FEC for 2.5-Gbps and 10-Gbps transmission, is developing a FEC chip that will support a 40-Gbps signal.
FEC, these companies say, has its uses in both long-haul and metro networks. For instance, Tony Stelliga, general manager of strategic marketing and business planning for Intel's telecom component division, said the primary market for the company is the metro network. Being able to extend the signal even a few extra kilometers could make the difference in a service provider's ability to reach some customers, he said. In the long-haul market, where amplifiers are used at intervals to boost the optical signals, FEC could be used to triple the optical network's reach before a costly regeneration of optical signals becomes
necessary.
3
VDSL:
Phone Companies Go Hollywood
Thanks to VDSL (very-high-bit-rate DSL) service providers are branching out to offer not just Internet access but video service, too, using existing copper. Faster than ADSL, the still evolving VDSL technology tops out at 52 Mbps downstream in asymmetric mode, or 26 Mbps in symmetric mode--fast enough to deliver quality video service. Among the phone companies testing the delivery of video service through VDSL: Qwest, which is providing service to more than 50,000 subscribers in Phoenix; Bell Canada, which is delivering bundled service to apartment buildings and condominiums in Toronto; and Norway's Telenor, which said recently it would trial broadband TV service in Oslo.
Most providers are keeping their eyes on VDSL developments, and a number of them are seeking to shape the creation of standards. The Full Service VDSL Group, formed last May to create international standards, includes BT, France Telecom, NTT, Deutsche Telekom and Qwest, along with equipment vendors Lucent, Motorola, ViaGate Technologies, Metalink and NextLevel Communications. For now, the range of VDSL limits deployments. Only customers no more than 1000 feet from the VDSL equipment attain the top speeds, while customers the farthest possible distance away, around 4000 feet, get service at perhaps 13 Mbps. Service providers offering VDSL typically extend fiber out to neighborhood nodes, where the VDSL equipment is installed, then deliver the service to MDUs (multiple dwelling units), where a high concentration of customers can compensate for high equipment costs, according to market research firm Cahners In-Stat. With the development of standards, equipment costs are likely to drop, permitting more extensive service deployment.
Unlike the Blockbuster-Enron initiative announced last year to provide video service over ADSL, video services delivered over VDSL are intended as a direct alternative to cable TV. The incentive for some customers to make the switch might be lower cost. In less urban areas, where cable companies might be slower to upgrade infrastructure, VDSL provides digital service that customers otherwise couldn't get. Michelle Abraham, In-Stat's senior analyst of multimedia broadband, notes ironically that, with the coming of VDSL, "Your best choice between DSL and cable might be to buy telephone service from your cable company, video service from your telephone company, and Internet access from one or the other." The market for video over DSL now stands at roughly 100,000 to 200,000 total subscribers, Abraham estimates--a miniscule amount compared with cable and satellite TV.
4
Free-Space Optics:
Last-Mile Alternative
Fiber-optic technology is hot, and wireless is hot. Combine the two and what do you get? Free-space optics (FSO): lasers beaming signals through the air rather than over fiber. Today, many businesses and organizations are unserved by fiber but need more capacity than T1 or T3 lines can offer. FSO provides an alternative to fixed wireless and could potentially offer bandwidth of up to 1 Gbps.
Compared with fiber-optic lines, FSO systems take much less time to deploy and cost much less--perhaps one-tenth to one-third the $100,000 to $300,000 cost of bringing fiber to a building, according to vendor AirFiber. But like fixed wireless, FSO is subject to environmental interference. While fixed wireless networks must account for rain fade, FSO technology must compensate for fog. Limiting the distance between FSO connections to a few city blocks permits systems to overcome signal diffusion and phenomena such as building sway, which can cause systems to go out of alignment. Because such technological issues have not been fully resolved, a number of FSO equipment vendors promise 99.999 reliability but advocate use of a backup data connection--a DS3 or even a fixed wireless system--just in case.
Thus, fixed wireless and FSO technology can be complementary. Strategis Group analyst Peter Jarich says XO, Teligent and other providers of point-to-multipoint fixed wireless services are considering the use of FSO for backhaul transmissions, i.e., for delivering bandwidth to the system hub, generally the largest building within a service area. One advantage to FSO: It requires no licensing, which cuts costs and speeds deployment.
FSO is already in use in the enterprise and MDU/MTU markets. The Smithsonian Institute in Washington, D.C., for example, uses FSO equipment from LightPointe Communications operating at 155 Mbps to create a high-bandwidth LAN among museums on the National Mall and administrative offices in nearby L'Enfant Plaza. The Four Seasons Hotel in Seattle uses FSO equipment from Terabeam to provide guests with data connections as fast as 100 Mbps.
Other makers of FSO equipment include Tellaire, LSA Communications, Optical Access, fSona and Canon. While Terabeam is attempting to provide service and sell equipment, commercial offerings of FSO are rare so far. Japan's DDI, Teligent, Allied Riser and XO are among companies that have tested it, but FSO as a service is only now taking off, according Jarich. The Strategis Group predicts worldwide revenue from FSO equipment will rise from $100 million last year to $2 billion within five years. "Everyone's waiting for fiber," Jarich says. "In some cases, carriers might never have the economic incentive to install it, so the market potential out there is big."
5
RPR:
Making Networks Roomier
Get more for your money, say advocates of RPR (Resilient Packet Ring) technology. Unlike SONET, RPR makes use of both rings in fiber-optic networks, letting carriers use and provision bandwidth more quickly and efficiently. In case of failure, the most important transmissions get sent first. Additionally, packets running over RPR networks are processed at their point of entry, with the ring acting as a kind of distributed switch that permits routers or switches to run at speeds faster than their throughput capacity. Service providers upgrading their networks may continue to use legacy equipment without affecting network performance.
RPR handles traffic up to eight times faster than SONET/SDH bidirectional line-switched ring or unidirectional path-switched ring systems for two reasons: Space freed by packets that left the network is immediately reused; packets are routed in whichever direction gets them to their destination the quickest.
Thanks to its ability to prioritize, RPR makes carrier-class services running over Ethernet in MANs possible, according to equipment vendors, and provides cheap transportation while delivering SONET-grade QoS. "Compared to Ethernet, SONET is quite expensive, so if you want to have all the good things about SONET, you port that over to RPR," said Maria Zeppetella, Probe Research's vice president of network infrastructure. "You get protection, scalability, carrier-class voice and video, and the link to legacy technology. This will allow Ethernet to be everything people have always promised it would be."
All major U.S. carriers and some CLECs are interested in the technology, if not already testing it, according to a spokesperson for Luminous Networks, which along with Cisco, Nortel, Dynarc and Lantern Communications is working on a RPR standard through the IEEE 802.17 Working Group. These companies along with Riverstone Networks, Avaya and others are part of the RPR Alliance, which is sponsoring interoperability testing. According to Riverstone CEO Romulus Pereira, products are available, but commercial use of RPR is at least 18 months away, given the time needed to complete standards and
testing.
While RPR is being designed to handle multiple services, many VoIP QoS issues remain. When first using RPR, carriers might run their data and relatively small portion of VoIP traffic over RPR and run the bulk of voice services over the same optical line using TDM over SONET, Pereira says. As VoIP technology improves, carriers could migrate the remainder of their voice traffic to RPR.
6
H.248/ Megaco:
Meeting of the Minds
The joint ratification of Megaco/ H.248 by the ITU and IETF has implications far beyond the protocol itself: It marks the first effort between telcocentric and Internetcentric organizations to advance the next-gen telecom network. It's a sign that VoIP is becoming a reality.
"From a service provider perspective it's very exciting to see a protocol like H.248, because we have been there since the development of MGCP (Media Gateway Control Protocol)," said Matt Johnson, senior manager of Level 3's Global Softswitch Division. "We like the fact that things we helped foster have now been embraced not only by the IETF but also the ITU."
Pressured to get workable products into the market, particularly after last August's interoperability event at the University of New Hampshire, many IP vendors have gone forward with MGCP-based products, although Megaco is being incorporated into them (e.g., Cisco's VSC3000, Nortel's Succession products; Santera's Santera One).
The common view among service providers and vendors is that Megaco will be used primarily as an intranetworking device-control protocol that will complement SIP's role transporting signals between networks. Many VoIP vendors are moving away from MGCP toward Megaco because it provides greater application-level support and is simpler to manage.
Many believe Megaco will become the official protocol to interface between external call agents known as the MGC (media gateway controller) and MG (media gateway). The MGC layer contains all media connections to and from the packet-based network and includes call-control intelligence and call-level features such as forward, transfer, conference and hold. It interfaces with those media connections, streams through signal applications and events and controls user-interface devices. This is all driven by the gateway control protocol, which performs master/slave control of the MG by the MGC to provide connection control, device control and device configuration.
Megaco is based on two primary concepts: termination and context. Terminations are media connections to and from the packet network that allow signals to be applied to media connections and events to be received from media connections to identify media flows or resources, implement signals and generate events. Contexts, defined as associations between connections and collections of terminations, act as a mixing bridge. The protocol can perform up to seven primary commands: add, subtract, move, modify, notify, audit and service change.
However, Jonathan Rosenberg, CTO of Dynamicsoft and co-author of the SIP protocol, argues that because Megaco places all device control within the network, the protocol is really a step backward. "This model is a bit controversial because it represents a more telcocentric way of building communication systems," he said. "Whereas the IP model calls for pushing intelligence to the edge, Megaco is much like the PSTN model in that all intelligence lives in the network."
7
Directory- Based Management:
Driving Service Awareness
The move to the next-gen service architecture has one dominant requirement: Know thy customer. Key to this evolution is the move to a network driven by a policy- and directory-based management system at the network edge. Until now, circuit-switched networks required manual provisioning by a service team for every network action, whether simple bandwidth upgrade or service allocation. With price points for basic services eroding sharply, a multiservice, on-demand network where both service providers and customers can control their service allocations will be key to survival.
With a policy-based network management tool in place, service providers can deliver SLAs and tiered bandwidth services tailored for specific customer needs, prioritize applications, and manage service demands. Customers receive one bill and have the ability to change a bandwidth allocation or access a new service on-demand. A number of well-established and up-and-coming vendors including Atreus, BroadJump, Cisco, Ellacoya, IP Highway, Sitara and Unisphere, are producing solutions.
Driving these systems are a number of emerging standard protocols such as COPS (Common Open Policy Server), LDAP (Lightweight Directory Application Protocol) and RADIUS. COPS is a stateful query and response protocol that can be used to exchange policy information between a policy server and its clients or policy enforcement points. LDAP is an Internet-based solution that keeps a directory of all user profiles, applications and network policies. RADIUS servers perform all initial service initiation functions, including authentication, authorization and accounting.
Unisphere's SSC (service selection center) portal uses COPS to communicate between a policy enforcement point and a decision point. It applies all policies to incoming traffic on its ERX-700 or ERX-1400 service switch and communicates through the COPS protocol with an LDAP-based directory to obtain necessary subscriber authentication functions. BroadJump's Control Works service management software products and their profiles are placed into an LDAP directory; subscriber authentication is performed using RADIUS; COPS is used to communicate with network elements and other policy control points. Allot Communications' NetEnforcer NetPolicy uses LDAP for accessing customer information from the customer directory and COPS to enforce SLAs.
Ellacoya's SGS (service generation switch) consists of three primary elements: hardware service generation switches, service creation manager and a business logic server. Each speaks to LDAP to view service, switches, prices, subscribers, applications and service context. The system also supports RADIUS as a plug-in to the business logic server for legacy subscriber management and start/stop session billing.
No matter which vendor a service provider chooses, production of these solutions is a sign that providers are becoming proactive. "It's all about personalization, whether that be the application, user or a combination of both," said Deb Mielke, principal at Treillage Network Strategies. "If a service provider knows my needs [through directory/policy and links to applications] it should be able to sell me more and control my experience."
8
Adaptive Modulation: Optimizing the Link
Because link conditions in a wireless environment change continually from one subscriber to another moment by moment, there is a need to tune and adjust dynamically to every subscriber on the network. While first-gen wireless networks sent the same signal to every subscriber regardless of their SLAs, second-gen wireless systems with non-line-of-sight characteristics must adjust to unpredictable conditions on each link. Enter adaptive modulation.
"Wireless is a challenging environment due to varying channel conditions, interference and high error rates," said Andy Fuertes, an Allied Business Intelligence analyst. "The wireless industry has put forth MACs and physical layers optimized for wireless channel conditions, including mechanisms for adapting to multipath and fading, error correction and retransmission. Adaptive modulation is also frequently offered."
Adaptive modulation can adjust the signal dynamically by varying the modulation schemes and changing modulation rates to deal with environmental and interference fluctuations between base station and subscriber. The system continually checks the error rate and sets the optimal modulation scheme for each subscriber. For example, a system could switch from 16 QAM (quadrature amplitude modulation), which has a higher bit rate but less interference, to QPSK (quadrature phased shifted key), which has less throughput but more immunity to interference.
A number of vendors are beginning to use adaptive modulation, including Aperto, Adaptive Broadband, Hybrid, Raze Technologies and Wiman (single-carrier modulation backers); BeamReach, Cisco and Iospan Wireless (OFDM-based companies); and Vyyo (wireless DOCSIS).
To optimize the link, vendors also are using other MAC and PHY layer technologies, such as adaptive FEC and adaptive power control. Dynamic FEC redundantly encodes information in a data stream to correct errors at the receiving end that may occur during transmission. It can determine and apply the best FEC level for each subscriber. Because subscriber signal-power levels vary, dynamic power control sets the right amount of power that should be sent when the signal is received without bleeding power into adjacent cells.
"Using things like adaptive modulation, adaptive FEC, antenna diversity and ARQ (adaptive automatic retransmission) can establish a hard brick wall to prevent jitter and latency and guarantee different classes of service," said Reza Ahy, president and CEO of Aperto Networks.
9
Raman Amplifiers:
Keeping the Light
The on-going optical marathon reached a remarkable milestone last February when the Williams network, using Corvis equipment, was able to transmit a signal 6400 km without electrical regeneration. A good part of this accomplishment was due to the use of Raman amplifiers. Shyam Jha, Corvis' vice president of marketing, predicts that Raman amplifiers will account for about half the optical amplifier market in two years.
The EDFAs (erbium doped fiber amplifiers) currently popular in networks boost the signal every so many tens of kilometers before an expensive OEO (optical-electrical-optical) conversion becomes necessary every 500 km. EDFAs, however, work in the C band, which most DWDM systems currently support, and also the L band, the expansion band many systems support to provide more channels. Thus in contrast to EDFAs, which operate in the relatively small 1525-nm to 1620-nm window, Raman amplifiers can operate in the 1300-nm to 1700-nm range (i.e., C, L, S, L-plus and S-plus bands) and probably over a much broader spectrum. Not surprisingly, a study by the Canadian Imperial Bank of Commerce (CIBC) estimated that the Raman amplifier market will grow from $220 million last year to $4.5 billion in 2005.
Raman, a concept that's been around a few decades, is complex, and the CIBC forecast was specifically about distributed Raman amplifiers. Distributed Raman amplifies the signal in the line fiber itself by pumping energy in the direction from which the traffic is coming. Discrete Raman is more akin to EDFAs because it uses a special narrow core fiber within the amplifier itself to boost the signal.
Regardless of which technology a carrier chooses, there are tradeoffs. Distributed Raman provides about a 6db gain. Discrete Raman provides an EDFA-like 30db gain, which affects how amplifiers are spaced along the optical highway: Each time a signal is boosted, so is noise, the nemesis of an all-optical network. While vendors such as Corvis and Nortel (through acquisition of ultralong-haul DWDM player Qtera) have chosen the distributed Raman route, Allen, Texas-based start-up Xtera Communications is taking a neutral position. Xtera recently launched a distributed Raman amplifier but plans to offer a combo discrete/distributed Raman product early next year.
10
SCTP:
SS7 over IP
Probably no other subject has been talked to death in the industry like VoIP. But while the reality of replacing the PSTN with a packet network remains for the most part in the future, there is now considerable momentum for moving a key component of the PSTN, namely the SS7 network, to IP. The impetus for this is an effort by the IETF's Sigtran (signaling transport) working group that released SCTP (Stream Control Transmission Protocol) as RFC 2960 last year. In fact, a few months before SCTP became standardized, a dozen companies, including Alcatel, Motorola, Nokia, Nortel, Sun Microsystems, Telcordia and Siemens, successfully completed interoperability testing of SCTP at a Siemens test facility in Munich, Germany.
Although SCTP is nonproprietary, the concept of running SS7 traffic over IP is not. Many companies in the Class 4/5 switch replacement business, such as Convergent Networks, have their own schemes for SS7 over IP. "Convergent's been running SS7 messages over IP for the better part of two years," said Seng-Poh Lee, the company's managing director for corporate applications engineering. "SCTP is not the only protocol out there, but if there's widespread support for it, then we too will support it." A vendor such as Convergent has the advantage of supplying all three key pieces of the VoIP puzzle--the media gateway, the media gateway controller (also called softswitch) and the signaling gateway--and can do its own thing. However another company, say one that supplies only the softswitch or the signaling gateway, might prefer to have a protocol blessed by the IETF, since SCTP is intended to be used for communications between the signaling gateway and the media gateway controller or media gateway, or between the media gateway and the media gateway controller.
While TCP also provides reliable data transfer, it was deemed too restrictive. Thus in an IP network, SCTP is intended to be used in place of TCP as a transport layer protocol. Vivek Pachaury, Hughes Software Systems' product manager for VoIP, points out that SCTP is intended to facilitate the transport of many PSTN signaling protocols, including SS7, ISDN (Q.931), and GR-303, in an IP network. Hughes claims to be one of the first to commercialize SCTP and counts Lucent and Motorola among its customers.
Meanwhile, TALI (transport adapter layer interface), a protocol that earlier was positioned to do pretty much what SCTP does, is now being cast by its developer Tekelec as complementary to SCTP because it would use the services of SCTP. Today, customers using TALI, which had a headstart on SCTP, include Level 3, Qwest and softswitch vendor ipVerse. SCTP's advantage, however, may be in having the imprimatur of the IETF. *
How We Decide What's Hot
and What's Not
We set down these guidelines for qualifying as a hot technology:
Hot means capable of entering the mainstream of telecom systems/operations within one to two years, while now having sufficient development dollars and industry support to become economically viable.
Telecom is defined broadly to encompass cable, interactive media and other emerging forms of communications, as well as the standard inclusion of voice, data and video transmissions.
A technology is not a product, but a product can execute a technology. To qualify, a new technology also must have the potential to make an impact on existing systems, operations or procedures.
2000's 10 Hottest Technologies
Soliton
Passive Optical Networking
VoDSL
Time Division Duplexing
Tunable Lasers
ODSI (optical domain service interconnect)
Synchronization
IP VPNs
10-Gigabit Ethernet
Very-Short Range Optics
Senior Technology Editor Sam Masud, and Staff Editors Sean Buckley and Ted McKenna contributed to this report.
RSNo. 305 |
