The venerable transport technology, synchronous optical network (SONET), has undergone a transformation in recent years as vendors try to make it more friendly to bursty, multiprotocol data streams without sacrificing the low cost, interoperability and protective features that have made it so valuable to carriers.
A new batch of startups is now planning to wave a magic wand over SONET to introduce even more data capability and efficiency in carrying multiple data types over SONET. Many of these companies will introduce product early in the year 2000.
"The key thing that they are all trying to do is to figure out how to deliver non-TDM (time-division multiplexing) services efficiently," says Mat Steinberg, optical networking analyst with RHK Inc., a South San Francisco-based consultancy. "If you want a T1 or a DS-3, SONET does a wonderful job of taking that, muxing it, protecting it and doing all the things SONET is designed to do." However, he says, most T1s are not fully utilized and the carrier wastes bandwidth provisioning them. "So what vendors are looking to do is what I call mining fiber. These vendors have different techniques to use the fiber more efficiently."
SONET (called synchronous digital hierarchy, or SDH, outside North America) has been the basis of transport technology for carrier optical networks. Designed originally to transport large volumes of voice circuits, it is based on a hierarchy of multiplexes of voice circuits that are channels with specific fixed bandwidths. These channels range from 1.5 megabits per second (mbps) (DS-1) to 51mbps (STS-1) to 155mbps (OC-3), 622mbps (OC-12), 2.5 gigabits per second (gbps) (OC-48) and 10gbps (OC-192). An OC-768 at 40gbps is planned for introduction in 2000 by several vendors.
SONET's strength has been that it is standardized and ubiquitous, and it works. Its weakness has been its fixed channels. Those channels are inefficient in transporting certain data protocols, such as Ethernet. For example, to carry 10mbps 10Base-T Ethernet over its SONET network, a carrier must allocate an entire 51mbps STS-1 channel, wasting 41 megabits of capacity.
This problem has had some industry observers predicting the death of SONET. Others feel the death knell of SONET is being sounded prematurely. "There are compelling reasons to use SONET interfaces," says Mike Guess, vice president of engineering, IXC Communications Inc., Austin, Texas. "That is why Cisco (Systems Inc., San Jose, Calif.) is still putting SONET interfaces on routers. What is really on the way out is high-level muxing with DS-1s and DS-3s, but it is still the best interface because it is so accepted." Because of interoperability, he adds, volume is driving down the costs of SONET chip sets, making the technology more affordable than ever.
"If vendors can play with that payload structure to make it fit a data world, there is no reason to throw away SONET with its protection features," Guess says.
New Tricks with SONET
That is exactly what vendors have tried to do over the past year. In fall 1998, a group of vendors founded the Data-Aware Transport Activity (D.A.T.A.) group led by ADC Telecommunications Inc., Minneapolis; Atmosphere Networks Inc., Cupertino, Calif.; Fujitsu Network Communications Inc., Richardson, Texas; and Nortel Networks, Richardson, Texas. The members of D.A.T.A. have agreed to an interoperable way to add asynchronous transfer mode (ATM) switching to their SONET add/drop multiplexer (ADM) equipment. This allows for switching and aggregation of data before it is put on a SONET mux--some call it statistical multiplexing of the flows. The integration of the feature into SONET equipment keeps costs reasonable.
Then in spring 1999, Cerent Corp., Petaluma, Calif., recently acquired by Cisco, unveiled its Cerent 454, which aggregates any voice, data or video service and concentrates it for transport across optical networks. Cerent eliminates the need to add new boxes for each new service. Instead, the Cerent box has general-purpose card slots that accept interfaces for muxes ranging from DS-1 to OC-192, as well as data types such as Ethernet, ATM and video.
"Cerent did a good job of engineering their box," says Steinberg. "They added an Ethernet switch card and ATM switch card, so they are able to do those functions right in the box, and it is all mapped back to an STS-1, or rather actually mapped to [the smallest unit of aggregation in SONET, which is any multiple of 155mbps]."
"They are being very pragmatic," says Guess. "They [have] all the best parts of SONET--the timing, performance monitoring and protection switching--and are reworking the framing structure so they can incorporate 10Base-T or 100Base-T and gigabit Ethernet, because that is what needs to be carried in the future."
If Cerent represents a second generation of SONET technology, companies such as Alidian Networks, Mountain View, Calif.; Luminous Networks Inc., San Jose, Calif.; and Siara Systems, Mountain View, Calif., represent a third generation. Their technologies typically involve aggregation of data traffic types and hand off to optical networks that maximize the capacity of the network and preserve the services characteristics of the data protocol, such as quality of service (QoS) in Internet protocol (IP).
Many of these third-generation products also support legacy TDM traffic and circuit-based services, in additional to data protocols such as IP and ATM. Some are putting router-like devices in their boxes, further integrating data functions with an optical network. Some are reducing costs and increasing functionality by using off-the-shelf chips and other components.
"They will take ATM, will take frame relay. They do mapping from packets and cells," says Steinberg. "So they are boxes that can take more than one service."
Meanwhile, Alidian will combine SONET reliability with the scalability of dense wavelength-division multiplexing (DWDM), adding service awareness features not available in either technology. Alidian will not aggregate electrical services, but it will take the output of that equipment, such as a digital subscriber line access multiplexer (DSLAM), DS-3 or OC-3, and aggregate it into flows ready for a DWDM wave. The difference is that, rather than devoting a particular wave to a specific service, such as an OC-48 wave for gigabit Ethernet services, Alidian can aggregate multiple protocols--IP, ATM, legacy SONET voice, fibre channel--in the same wave. Alidian also can carry these services in native mode.
As for Siara, that company will base its technology on providing IP services in access networks, but will provide those services alongside legacy services and within SONET transport. The company has developed proprietary silicon, including application specific integrated circuits (ASICs) for packet processing and SONET system-on-a-chip. Siara is seeking to tailor its technology for service providers who are integrating vertically, providing service from the last mile area to international networks, such as AT&T Corp.; GTE Corp., Stamford, Conn.; Level 3 Communications Inc., Omaha, Neb.; MCI WorldCom Inc.; Qwest Communications International Inc., Denver; and Sprint Corp.
Although its technology is designed to hand off to SONET systems, Luminous believes it can provide transport in access optical networks using IP natively on fiber. Wayne Williams, vice president of marketing, says the company's key intellectual property is a solution to the timing problems to putting IP directly onto glass, and it will work within a controlled ring with Luminous' own ADMs. Luminous will hand off traffic to end users or access networks using T1 or DS-3 interfaces and will also hand traffic up to core networks through standard SONET interfaces.
Despite the many attractive features of the new technologies, these products are largely unproven and a killer application has yet to emerge. At the same time, fiber is plentiful in most systems, and SONET prices are dropping. Add to that the fact that more products are waiting in the wings and it is unlikely service providers will make any early commitments. "They see there is new product coming out and it is beneficial to wait. Also, the software is way behind," Steinberg says.
Traditional SONET vendors also have added new wrinkles to their equipment. Albert Ng, vice president of engineering at competitive carrier e.spire Communications Inc., Annapolis Junction, Md., says much of e.spire's traffic is 10Base-T or 100Base-T Ethernet, neither of which is a good fit with a SONET mux. Rather than waste bandwidth, e.spire is working with regular supplier Fujitsu using a product called Fast Lane, which aggregates four 10Base-Ts for an STS-1.
e.spire would like a similar interface for 100Base-T, where the company sees increasing demand. "It is one of the most common applications for our customers," says Ng. "We see the bigger corporations migrating to 100Base-T."
100Base-T usually runs at only 50 percent or 60 percent of capacity to avoid packet collision problems, so e.spire is looking at putting two 100Base-Ts in an OC-3. While there would still be some bandwidth waste, it would be acceptable, Ng says.
NeoSONET Services
Carriers are struggling to juggle the bandwidth demands of data and, in the future, will need to add specialized services for data. For these challenges, "neoSONET" technologies may provide solutions.
"We scheduled a meeting in December with Cerent to go over that technology," says Ng. e.spire expects to use the technology "in a metro environment where we are interconnecting with customers with very high-speed intranets, and also for our long-haul network. With gigabit routers coming out, we think that intranets or extranets are going to need higher and higher speeds for transparent LAN (local area network) transmission and may need that type of equipment."
Others are not as convinced the technology will work for them. Randolph Nicklas, director of data architecture, NEXTLINK Communications Inc., Bellevue, Wash., says, "It remains to be seen what the price points will be and what total system costs will be, not just initial costs, but also ongoing costs. We know how to run SONET rings in the metro space, so if the new players in the game can just push down the cost of traditional SONET gear, that may be at least as or more attractive as new equipment. We are not going to jump on someone who comes out with standard muxing to haul things back to our ADM point of presence (PoP)."
