Is Ethernet the ticket for 'trial and wait' PON customers?

TRENDS

By KATHLEEN RICHARDS

The debate over which protocol is best for framing traffic in passive optical networks (PONs) may boil down to this: a promise of lower costs for more bandwidth versus a lack of standards.

After making its way out of the enterprise into metro networks, Ethernet is now emerging as the framing technology in PON solutions from several optical access startups-Alloptic Inc., OnePath Networks, and Salira Optical Network Systems, for example. Ethernet makes sense, according to these vendors, because of its advantages over ATM, the cell-based data protocol recommended for use in PONs by the Full Service Access Network (FSAN) coalition of 21 network operators and about 30 vendors-a group initially formed by network operators to discuss their access networking needs. FSAN helped develop the International Telecommunication Union-Telecommunication (ITU-T) ATM PON (APON) standard G.983.1.

What are Ethernet's advantages? Lower costs, scalability beyond APON's current 622.08 Mbits/sec, and simplicity-Ethernet is widely used and understood among business customers. Currently, in ATM PON configurations that conform to the ITU-T standard, as many as 32 customers may share 622 Mbits/sec downstream and 155.52 Mbits/sec upstream-certainly not the level of bandwidth that business customers expect from a fiber connection.

The FSAN coalition is working on Gigabit PON (GPON), a proposal to increase the bandwidth specified in the ITU-T G.983.1 standard from 622 Mbits/sec to 1.2 Gbits/sec. Some in the ATM camp also disagree with the general assumption that Ethernet framing offers cost and scalability advantages. "In the optical domain, cost would not be determined by the protocol efficiency but by the desire to cover certain optical specifications that include power level, distance, split ratio, and the spacing between bursts that go upstream," asserts Boris Auerbuch, chief technology officer at APON vendor Terawave Communications (Hayward, CA). "That's the fundamental cost tradeoff for the component selection."

"With PONs, the underlying communications is TDM downstream; it is at the framing layer that you get the differentiation," counters Herbert Martin, president and CEO of Salira Optical Network Systems (Santa Clara, CA). "We can handle many, many more users. Ethernet is a plug-and-play technology. There is no need to do provisioning each time, which makes it much less expensive."

Salira's product will enable carriers to support native TDM traffic and use their existing operations support systems and test procedures. "If you are going through a traditional Ethernet interface, you are doing circuit emulation; you convert the voice into an Ethernet frame," explains Jim Diestel, director of product marketing at Salira. "The thing that we are doing that is unique is that we are leaving the TDM services in their native mode and just putting T1 services on fiber. The advantage is that the support systems...continue to work, they would just be reporting what is happening through the fiber portion and the electrical tail circuits. It then takes the number of repeaters that are necessary in the network to extend copper for long distances out of the equation, so costs are reduced."

Salira is primarily marketing its Ethernet PON (EPON) equipment to carriers servicing the small and medium enterprise customer. The company has disclosed generalized information about its architecture, which is designed to support TDM and packet-based services, and expects its products to be ready for lab evaluations sometime next month. Salira's EPON offering also addresses common Ethernet sticking points such as a lack of quality-of-service (QoS) capabilities and security. ATM readily handles such concerns, which is why ATM is entrenched in many carriers' networks. With Salira's PON solution, QoS is realized using a combination of VLAN 802.1q and the ability on a 10/100 Ethernet port to set the latency rate and control the availability rate through that port into the PON.

As EPON offerings make their way into lab tests and network trials, high-profile APON vendors such as Quantum Bridge Communications (Andover, MA) and Terawave report trials and contracts with a range of customers.

Quantum Bridge services competitive local-exchange carrier (LEC) customers such as Comcast Business Communications (Morris, NJ), municipalities, and broadband utility Sigecom (Evansville, IN), among others. The company looked at an Ethernet approach as early as 1999, but chose instead to pursue APON because of the FSAN initiated standards. "We've been very involved in the FSAN standard along with the large carriers, the [incumbent] LECs, and the MSOs [multiple system operators] who have a lot of ATM in their network," says Charlie Guyer, director of corporate communications at Quantum Bridge. "A lot of these guys are very concerned about quality of service. If they are attacking the business market, they need to lead with either voice or data, and voice has got to be reliable; it can't be a science project as some of the voice-over-IP solutions are."

Despite its APON successes, Quantum Bridge is focused on customer solutions, not "protocol holy wars," says Guyer. Thus, the company also offers an "Ethernet in, Gigabit Ethernet out" approach, accomplished using an Ethernet interface card on the customer premises system and a Gigabit Ethernet WAN switch card in the central-office equipment. This technology was announced in February and is currently in use in the Sigecom network.

The company will bring a standards-compliant EPON system to market when standards are developed and mature, acknowledges Guyer. Quantum Bridge has been elected to serve as the liaison between the FSAN coalition and the IEEE 802.3 Ethernet in the First Mile (EFM) subcommittee, which is working on developing a physical layer standard for Ethernet in access networks. Frank Effenberger, director of systems engineering at Quantum (previously with Telcordia Technologies), is writing the optical baseline specifications for EPON in the EFM group.

The IEEE 802.3 EFM subcommittee is working to define several aspects of EPON over a single strand of singlemode fiber. The EFM standard will also cover point-to-point Gigabit Ethernet and point-to-point copper specifications. The EPON issues currently being debated by industry representatives from service providers and vendors include the number of subscribers (or splits) the standard should support, the distance, the protocol used to manage how the PON transmits and receives data, and how the PON arbitrates access among the attached subscribers. Operations, maintenance, and administration issues are also under evaluation.

After reviewing numerous technical proposals, the EFM subcommittee is expecting to select the baseline proposals included in the standard sometime this spring. The first working ballot is scheduled for November 2002 and ratification of the standard is expected in September 2003.

A few points of contention have surfaced among carriers and the enterprise contingents involved in the EFM standards discussions. "Carriers want range," says Salira's Diestel. "They are looking at a 20-km range to maximize the split ratio to 32:1, especially for residential networks. From an enterprise perspective, there isn't as much interest in the higher split ratio."

The 802.3 EFM standard is a start, but it will not include any specifications for services other than Ethernet. "After this ratification, they need to go a step further to try to do some more standardization to make this happen," says Terawave's Auerbuch. "The bottom line is if you are trying to deliver mixed or multiple services reliably, essentially standards-based, then APON is the answer. If you are trying to deliver data applications only, then EPON might simplify it, because then you don't need to deal with the ATM provisioning."

Meanwhile, the FSAN coalition is debating the structure requirements for the GPON architecture. These discussions started last September. According to Terawave's Auerbuch, the protocol will be cell-based and allow variable-sized bursts upstream. After FSAN develops and finalizes its standards proposals, the coalition will make recommendations to the ITU-T.

"When we started the company, there was no APON or EPON, there was just PON," says Quantum Bridge's Guyer. "Now people are trying to divide it up and paint one as more beneficial than the other. But right now, people have bandwidth needs; carriers want to offer new services. Before this whole downturn, people had funded the backbone, they had funded the metro, and those started to dry up a little bit. The next obvious place for investment is going to be the access networks."

Venture capitalists and other investors apparently agree, as multiple optical access vendors specializing in PONs received funding during the last two years. "Optical, which today is dominantly point-to-point, is basically maturing and reaching point-to-multi point connectivity," says Auerbuch. "Nobody is saying that the future will be PON in the proprietary format. The trend is to look to standardize the solution that will allow at least the physical design of the network, optics, and cable, to be the same from one equipment vendor to another."

More in PON/Systems