PONs: Passive Aggression

Availability of fiber: The shortage of fiber in access nets means that PONs can't be used everywhere.

Distance limitations: As PONs are passive - there's no electrical amplification or regeneration of light signals - their range is limited.

Shared-media issues: Like cable modems, PONs share bandwidth among users, limiting scalability.

Product availability: At present, only one vendor --Optical Solutions Inc. http://www.opticalsolutions.com -- is actually shipping PON products.

Missing details: Many of the vendors provide skimpy information on their products, which often makes it tough to tell whether they will address key carrier requirements, such as security and reliability.

A build-your-own matrix giving a detailed competitive analysis of PON equipment providers, on Table 2: Business PON Vendors

An interactive design tool that gives general guidance on the applicability of PON systems - how many sites a single access line can support, at different bandwidths and distances. Play with it on /interactive/pon1.asp.

Guidance notes that puts PONs into context, take a close the technology's potential and limitations, and point to importance issues coming out of our competitive analysis. You can read these notes sequentially, or use the following hypertext links to jump straight to particular issues:

A PON Primer
The Case for PONs
PON Cons
PON Products Today
PON Architectures
PON Capacity
Speeds and Feeds
Reliability Risks
Alternatives to PONs
In a sense, the term passive optical network is a misnomer, since the only passive elements of a PON are the splitters located in the outside plant between central office and customer sites.

At either end of the PON are very active devices engaged in generating and controlling light signals. Here's a more detailed breakout of the main elements of a PON, followed by bullet points on how they work.

Central office: To provision services over a typical PON, a carrier requires a special switch called an Optical Line Terminal (OLT).

The OLT's job is to send traffic downstream to network subscribers and handle its upstream return from subscribers. Typically, traffic is sent in both directions at 155 Mbit/s. Upstream and downstream traffic use different frequencies on one fiber to avoid interference.

As already noted, downstream and upstream traffic are handled in different ways:

Outside plant: PONs rely on passive optical splitters placed inside controlled environmental vaults in manholes, under the curb, or in ruggedized outdoor cabinets near office parks or buildings. As the light broadcast from the OLT hits the splitter, it is deflected onto multiple fiber connections, depending on the splitter used. Splitters feature two to 32 branches, but it's important to note that each split consumes precious power as well as cutting a portion of bandwidth available. Splitters can be positioned to create PON star, ring, or tree configurations.

Customer premises: PON networks terminate in so-called optical network units (ONUs), also known as optical network terminations (ONTs). Their job is to take in light that's sent from the passive splitters, convert it to specific types of bandwidth (such as 10/100-Mbit/s Ethernet, ATM, and T1 voice and data), and pass it on to routers, PBXs, switches, and other enterprise networking gear. ONUs also incorporate the lasers that send traffic back to the central office at the command of the OLT.

ONUs can be installed directly in the wiring closet or data center of the customer premises. But for many carriers, it will be more efficient to put them in outside plant locations, so that customers can hook into the PON from DSL (digital subscriber line) services. This approach gives customers the benefit of optical networking without new fiber. It also enables carriers to offer PON service under existing DSL tariffs.

Standards

PONs emerged from the lab thanks to a 1995 group formed by British Telecommunications PLC http://www.bt.com. Dubbed the Full Service Access Network (FSAN, http://www.fsanet.net) coalition, the group soon included Alcatel, BellSouth, Fujitsu, Lucent, NEC, NTT SBC, Siemens, Nortel Networks, NTT, and a range of other carriers and equipment vendors.

The goal of the FSAN group was the same as its name: to find the cheapest, fastest way to create a "full service access network" that would extend emerging high-speed services, such as IP data, video, and 10/100 Ethernet, over fiber to residential and business customers worldwide.

The group decided to use ATM (asynchronous transfer mode) over a simple physical network with a minimum of moving parts. ATM is an ideal transport for PONs because it supports quality of service guarantees for a range of traffic types-voice, data, and video--over a single link.

By 1999, specs for a fundamental ATM PON had been written and approved by the ITU-T (International Telecommunication Union, www.itu.org) as specs G.983.1 and G.983.2 in the ITU-T's Study Group 15 (which addresses transport networks, systems, and equipment).

This specification helped fuel the marked increase in PON product development that's taken place over the past six months. Today, trials of commercial PON gear are underway around the world, including projects sponsored by NTT in Japan; Bell Atlantic, BellSouth, Comcast and SBC in the U.S.; and Singapore Telecom.

Because PONs are so new, their exact role in today's networks hasn't been established. Still, many carriers have clear plans about the advantages of deploying them. Here are the key benefits they cite:

Fiber extension. PONs can extend the use of fiber to business customers located in the "last mile." And there are plenty of these customers. According to Vertical Systems Group Inc. http://www.verticalsystems.com, a market research and consulting firm, 76 percent of all U.S. businesses in 1999 were located within a mile of an available source of fiber--with no link set up to get at it.

PONs let carriers serve these customers quickly, without adding miles of new fiber, because one strand hooked to a splitter serves multiple sites. "PONs reduce costs by giving us a way to share objects and fiber in the central office and distribution network among multiple customers," says Dan Spears, research director for BellSouth Science and Technology http://www. bellsouth.com.

Ease of provisioning. PONs offer a way around modifying the Sonet infrastructure. In Sonet networks, bandwidth is issued to customers by assigning them time slots on redundant rings. Trained personnel must coordinate settings among central-office cross-connects, add/drop multiplexers throughout the ring, and CPE. This is a labor-intensive proposition. PONs, on the other hand, allow multiple fiber connections to be provisioned from a single connection in one location.

Bandwidth flexibility and leased-line consolidation. Sonet bandwidth is doled out rigidly at specific rates, and the increments between these rates vary widely. (Consider the jump from Sonet OC-3, for example, at 155 Mbit/s, to OC-12 at 622 Mbit/s.) Since fluctuations on corporate Web sites and intranets often make it tough to predict bandwidth requirements, a customer needing a bit more bandwidth on a Sonet ring can force a costly carrier upgrade that largely goes unused. In contrast, PONs allow bandwidth to be allocated quickly and easily at a range of rates. All that's needed is another splitter added or taken away.

PONs also work well for customers looking to simplify their leased-line setups. PONs can be configured to give exact increments of bandwidth needed, instead of forcing a customer to buy multiple T1s/E1s along with fractional services. And PONs don't require the extra configuration and management of multiple leased lines.

Reduced cost. It's not unusual for a Sonet ring upgrade involving the change of several nodes on a ring from OC-12 to OC-48 to take weeks. It can also cost a carrier $500,000 or more, when engineering help, equipment, and new fiber is finally tallied. With a PON, the reduced time, effort, and equipment required result in a fraction of that figure. Early product data indicate about $5,000 to $7,000 per customer. For a PON with a maximum of 32 customers, that's still a savings at $224,000.

Alternative to HDSL. PONs can give xDSL customers the performance advantages of fiber at higher data rates. "HDSL (high-speed digital subscriber line) is a noisy signal that interferes with other services," says Alain Granger, director of product marketing at Alcatel USA http://www.alcatel.com. In contrast, he says PON trials improve performance with data rates of 155-Mbit/s in both upstream and downstream directions. (HDSL is generally limited to 1.544 Mbit/s in both directions without repeaters.) PONs also are ten times faster than the top-speed DSL service, ADSL, which runs at about 8 Mbit/s downstream and 640 kbit/s upstream.

While PONs offer advantages over Sonet and xDSL, it's not likely they will replace them anytime soon. This point is clearly made by both carriers and PON vendors, who are setting up equipment and services wherein PONs work alongside other services. Lucent and NEC, for instance, plan to offer PON products that are really add-ons to existing ATM switches that also support a range of other services.

Before they become established alternatives to anything, PONs will have to prove their worth in live networks. And doing that will require PON vendors to show they can cope adequately with the hurdles of PON technology. Read on. Despite their advantages, PONs face significant obstacles on the road to success, as noted in our introduction. Two of these challenges are technological and present the toughest barriers to PON deployment. Let's look at them in depth.

Shared media issues. First, the good news: The fact that PONs share bandwidth among multiple subscribers lowers service costs and helps carriers efficiently amortize the equipment and operations expenses.

And the bad news? Any amount of upstream bandwidth transmitted over a PON will be divvied up among the number of users at the end of the fanout (the number of fibers that wind up terminated by ONUs on customer sites). Therefore, on a 155-Mbit/s PON link with four splits, each subscriber will receive 38.75 Mbit/s.

That's not a bad amount of bandwidth. But on many PONs, splitters are added to links that already have been split. And that means bandwidth is quickly consumed. Add an eight-point splitter, for instance, to our 38.75-Mbit/s link, and each user is guaranteed just 4.84 Mbit/s of bandwidth.

Distance limitations. The fact that PONs don't regenerate or convert optical signals midnetwork makes them cheaper, but it also limits their reach. Without regeneration, light signals lose power quickly, consequently losing transmission capability.

According to the FSAN specs outlined by the ITU, PONs have a theoretical distance limitation of 20 kilometers (roughly 12 miles). This distance is required, FSAN says, for a PON to exercise its so-called ranging protocol. That is the protocol used by the OLT to ensure that the ONUs at the customer site will all be able to transmit bits from their multiple endpoints back to the OLT without collisions.

In fact, 20 km is a long way from reality for many PONs. The actual distance depends on the power of the laser used to transmit the light, and the reductions in power that the light suffers along the way.

This "power budget" calculation forms the basis of our interactive PON Design Tool, on /interactive/pon1.asp, which aims to give a general idea of PON performance.

In order to keep costs down, most vendors use the least expensive lasers in their PON equipment - ones with power levels of between 25 to 30 decibels. That's just enough to send a signal directly from the laser source 20 km to the customer site to accommodate the requirement of the ranging protocol.

Now start subtracting: PON OLTs, ONUs, cabling and splitters all take power out of the starting budget. How much power? Generally speaking, the connector from OLT to fiber takes an average of about .3 decibels from the strength of the initial signal. And it is not unusual for .27 decibels per kilometer to be robbed from the signal as the light travels at 155 Mbit/s down a strand of fiber.

Once splitters are added, more power is lost--typically, about 3.8 decibels for each two-way split.

The bottom line is that there's a trade off between distance, bandwidth and the number of sites supported by a single access line into the OLT. For instance, our design tool indicates that PON equipment could support six sites at 9.87 kilometers, delivering 25.8 Mbit/s each. If eight sites were configured, the distance would shorten to 8.6 kilometers and the bandwidth would go down to 19.37 Mbit/s each.

The Invisible MarketSeven vendors have PON products in various stages of availability. Of these seven, one is a European company, Alcatel SA http://www.alcatel.com. Three are subsidiaries of Japanese companies: Fujitsu Network Communications Inc. http://www.fujitsu.com, NEC Eluminant http://www.eluminant.com , and Paceon Corp. http://www.Paceon.com. (Paceon was founded by Mitsubishi Electric Company in January 2000.) Three are U.S.-based startups: Ignitus (now a part of Lucent http://www.lucent.com), Quantum Bridge http://www.quantumbridge.com , and Terawave http://www.terawave.com .

As noted earlier, few of these vendors have products to show. The vendors say a PON avalanche will occur late in 2000, with most products set to ship during the second and third quarters. What's more, many refuse to give specific details about their upcoming offerings-- a fact that's clearly shown in the competitive matrix that accompanies this report (see table 2). Most of the vendors say the gaps are in areas where they feel that to divulge information would compromise their competitive advantage.

In fact, skimpy information about PON products may indicate something much worse--a gap in functionality. When questioned about the level of security offered in their products, for instance, most of the vendors interviewed for Light Reading's report seemed unaware of what to say. After a long pause, most said they were "FSAN compatible." When asked specifically what that meant, none could give a ready answer.

Since the FSAN specs call for all transmissions on the PON to be encrypted, these vendors are covered. But in heavy-duty Intranet environments, customers require much more than this level of security. Only Lucent, Quantum Bridge, and Terawave claim to add password protection to the standard encryption provided by the FSAN specs.

But these claims are tough to verify. In fact, there is no way to verify the claims of most PON vendors to have any products at all. To a one, the vendors interviewed for this report failed to put us in touch with their trial customers, despite repeated requests. And when contacted, BellSouth, which we contacted because Lucent reports it as a trial customer for Lucent's business PON products, refused to confirm it. "We have sent an RFP to a fairly small group of suppliers and haven't announced the list of names. We don't want to alienate anyone," says BellSouth's Dan Spears.

This situation makes it tough to determine just how PONs will perform in the real world. And it signals that PONs are being shaped by feedback on features and functions that may exist only in kinky prototypes. Again, vendor claims about the capabilities of any PON product need to be verified on the proving ground of live tests and trials in carrier networks.

It's tough given these gaps to determine product differentiators for gear that's really nascent. Still, information available today indicates some key areas to watch. These areas are covered as follows, with information about how vendors are stacking up their offerings to support each one.

The basic elements of PONs are switches (OLTs) with nonblocking architectures, supporting aggregate rates of about 5 Gbit/s. There are three exceptions: Lucent says its GX 550 supports 25 Gbit/s. Paceon offers 20 Gbit/s capacity in its switch. Also, Terawave says it has two switch fabrics, one for ATM supporting 5 Gbit/s and one for TDM switching with a capacity of 20 Gbit/s.

For the most part, PON OLTs are dedicated to passive optical networking. But at least three PONs---Fujitsu's Flashlink and Lucent's GX 550 and Ignitus 3500 - are really PON blades designed as options for larger ATM switches The advantage of this is that carriers can derive multiple functions from the switch.

On the downside, carriers who want to benefit from the use of PONs from these vendors will have to already own their ATM switches, or be ready to invest in them.

Most PONs, whether they come with dedicated OLTs or consist of OLT blades, are based on ATM. As noted, FSAN chose ATM for its support of multimedia traffic, including voice, video, and IP data. Quantum Bridge, however, says its PON supports native IP connections in addition to ATM, not just IP-within-ATM as the other vendors do. If this is true, it means that Quantrum Bridge has achieved a breakthrough, since no other vendor can claim to support native IP in its PON. Unfortunately, however, Quantrum Bridge can't back its claims with customer testimonial.

Physical ConnectivityCapacity is one of the key features of emerging PONs. And the overall speed of the switch is only one indicator of the real size of network a PON product can handle.

It is also vital to check the number of physical ports that emanate from the PON's OLT. These vary from 40 ports (for the PON module that fits into Lucent's GX 550 ATM Switch) to just 7 (for Quantum Bridge's Optical Access System).

Each PON port is capable of supporting a 32 distinct endpoints--consisting of ONUs on or near multiple customer premises. Alcatel says it plans to offer a special version of its PON that will allow for 64 endpoints.

But port count doesn't tell the whole story about PON capacity. While most OLT ports control one two-way PON connection over a single fiber, there are exceptions. Quantum Bridge, for example, supports dual-fiber as well as single-fiber links on its OLT. Quantum Bridge says this means that it will be able to pack more links into each PON connection, because it will be able to support two upstream and two downstream connections instead of one up and one down.

DWDM ChannelsAnd some vendors claim they'll be packing more channels than others onto each fiber. Quantum Bridge and Lucent's Ignitus, for instance, say they can accept DWDM input from the carrier or WAN side of the OLT, increasing the input capacity of their PONs. Instead of supporting a single channel per connection, these PONs will be able to support several channels, by accepting input from DWDM multiplexers located next to the PON OLT in the central office.

But it's important to keep in mind that these PONs can accept just two, four, or eight DWDM channels per fiber from another source at the central office. They will probably not be able to work with multiplexers capable of providing more channels, because the low-powered PONs will not be able to keep up with input from more powerful gear.

Two vendors, Quantum Bridge and Terawave, say they implement DWDM on the customer side. In this case, the PON OLT will be able to link to DWDM muxes housed at customer locations, or inside points of presence. A CLEC, for instance, will be able to DWDM multiplex long-haul services it gets from another carrier, providing its customers with a range of services from one link. Also, Terawave says the use of customer-side DWDM will encourage CLECs to offer wavelength-only services, or lambda services.

PON vendors also are intent on dividing up lightwaves their own way. Terawave, for example, says its support of DWDM on the customer side is meant to work with its proprietary provisioning scheme, called SALSA, which lets carriers assign specific amounts of optical bandwidth on DWDM channels to specific users from a remote management console.

Quantum Bridge has a similar scheme, called Dynamic Wavelength Slicing, that lets carriers remotely allocate PON bandwidth--including bandwidth contained in the DWDM channels Quantum Bridge access from the WAN.

All PONs today support data rates of 155 Mbit/s downstream and upstream. But the FSAN specifications leave room for greater speeds. Today, only two, Terawave and Quantum Bridge, support 622 Mbit/s downstream and upstream. But no matter how fast it may be, a PON is only useful if it supports the kinds of services customers need. Today, since most PONs are being positioned to replace or consolidate leased lines, all products are capable of delivering T1/DS1 (1.544 Mbit/s) connectivity.

Terawave also supports fractional T1. Paceon and Terawave offer T3/DS3 (45 Mbit/s) services. Paceon and Quantum Bridge can feed Sonet OC-3 (155 Mbit/s) connections. Lucent and its Ignitus subsidiary and Quantum Bridge support ATM; and Terawave offers frame relay.

When it comes to LAN interfaces, all the vendors support 10/100 Ethernet to customers in their PONs. But only Terawave offers a Gigabit Ethernet link. One reason it can do this is that Terawave offers a neat little add-on called a PON Repeater that boosts the PON signal midnetwork. Call it cheating; but no doubt this tool will serve handily in designing real-world PONs. So far, Terawave is the only vendor to offer a "cheater" amplifier for PON attachment.

PONs need to be able to transmit bandwidth from other services on the WAN side. That's why most PONs support either OC-3 (155 Mbit/s) or OC-12 (622 Mbit/s) trunk connections. Lucent and Ignitus offer OC-48 (2.5 Gbit/s) because their PON platforms are really big ATM switches, for which PON blades have been designed.

It's interesting, if a bit alarming, to note the nearly complete absence of support for European interfaces in today's PON products. This gap is even stranger considering that so many European carriers, including BT, contributed to the FSAN specs. Only Lucent presently supports Sychronous Digital Hierarchy with STM-1 and STM-4 connectivity.

Customers seeking to replace leased-line services with PONs are deeply concerned with reliability. And they may get very worried when they see what PONs are offering in this regard.

Nearly all the vendors interviewed told Light Reading their PON reliability is based on automatic failover to manually configured redundant links.

This approach reduces the capacity of the PON, because it means that for each link to a customer ONU, a separate link will need to be set up to protect it. Also, there's no specific standard, as there is in Sonet, that guarantees the failover will occur in less than 50 milliseconds.

On the plus side, over half the products will ship with support for PON rings. This type of configuration allows carriers to set up mirrored rings that make redundancy faster and more efficient. Only Alcatel's APON, Fujitsu's Flashlink APON, NEC Eluminant's Fiberslam, and, surprisingly, Lucent's GX 550, won't have integral ring support in the first release.

Many questions about PON products remain unanswered. Because of this, it's logical for carriers to seek alternatives to PON deployment. And there are a growing number of such alternatives.

Vendors like Appian Communications http://www.appiancom.com, for instance, are coming to market with products designed to provide alternatives to Sonet delivery for metro optical networks (see Appian Teeters On the Edge ). Appian's first product, the QSAP 4800, grooms bandwidth in increments of 64 kbit/s to 1 Gbit/s onto metro networks and supports DWDM. It is slated to ship this summer at a starting price of $25,000. At prices like that, it's tough to make a case for a more limited PON.

And Appian's not alone. Other vendors, such as Alidian Inc. http://www.alidian.com are readying metro bandwidth delivery platforms. And as costs of other MSPPs come down, those products too will challenge the use of PONs (see Sonet Goes POP).

Another prominent alternative is offered by LuxN Inc. http://www.luxn.com, whose WavStation product is often compared to a PON because it is designed to extend carrier bandwidth to customer sites within the last mile of fiber.

The WavStation is a 16-slot chassis that accepts input from Sonet rings, routers, DWDM multiplexers, and other gear in the central office and converts it to OC-3, OC-12, OC-48, Gigabit Ethernet, fibre channel, and voice service. (The vendor says it plans to release an add-on card for the WavStation that supports 100Base-T Ethernet and four T1 connections early this summer.) At the customer premises, a device called a WavPortal accepts the signal and delivers it in the required interface to PBXs, switches, routers, and other gear.

The WavStation also compares to a PON because it does not use any active amplifiers to boost signals between the CO and the customer premises. This allows LuxN to keep its costs lower than those of Sonet equipment, metro DWDM gear, and MSPPs. A typical configuration starts at about $30,000.

Despite its similarity to PONs, however, the WavStation differs them in some important ways. It uses stronger laser signals, and it does not use splitters to deliver bandwidth to each customer. Stronger signals and no splitters allows the WavStation to send traffic greater distances than a PON (typically, 35 to 70 kilometers).

Clearly, LuxN's solution will work better for some carriers than a PON. But there is also a key drawback to LuxN's approach. The huge amount of bandwidth delivered to the customer premises requires an equally huge port on a router or switch. And port real estate isn't cheap-on high-end Cisco routers, they can be worth $50,000 or more apiece.

Still, for some carriers, the reliability and distance range of LuxN's solution will weigh favorably against its cost. And for some, this will surely offer an alternative to PONs.

Incidentally, LuxN isn't opposed to the PON approach and says it plans to support a PON extension to the WavStation sometime in the future.

LuxN's openness to PONs isn't unique. It's being taken by a range of other vendors, such as Adva Optical Networking Inc. http://www.advaoptical.com a DWDM vendor that offers a PON blade for its platform.

In fact, many observers say it's clear that PONs will be an ongoing part of solving the "last mile problem." They say PONs will be integrated with other approaches as needed to achieve economical solutions in the metro and residential access markets.

"PONs are part of an evolution. They're certainly not a complete solution," says Barry Moon of RHK. "When a new technology comes to the fore, it's easy to forget that other technologies are evolving as well. We're going to have a variety of different methods for deployment."