Introduction

Network service providers are investing aggressively in installing fiber cabling to premises, curbsides, and neighborhoods in efforts generically called “fiber-to-the-X” (FTTX). The goal is to bring as much bandwidth as possible as close to as many potential subscribers as possible. Carriers are making the investments to provide “triple-play” voice, data, and video services today and, more importantly, to assure that they will have the ability to create new, revenue-generating services in the future.

FTTX has been a long-cherished goal for carriers. The time division multiplexed-passive optical network (TDM-PON) architecture as we know it today-an economical method of sharing a single fiber and central-office (CO) transceiver among many customers-was conceived in the late 1980s, when FTTX concepts were first developed.

However, actual FTTX deployment has been very limited until recently. Now, as deployment finally begins in earnest, it makes sense to step back and question whether the basic TDM-PON design, conceived so long ago, is still the most appropriate network architecture for the present and the future. This discussion applies to all flavors of TDM-PONs: ATM-PON (A-PON), Broadband-PON (B-PON), Ethernet-PON (E-PON), and Gigabit-PON (G-PON).

Why Rethink the TDM-PON Architecture?

There are two key reasons to rethink the TDM-PON architecture:

1. It is questionable whether TDM-PONs will scale adequately to meet ever- increasing bandwidth expectations. A-PON/B-PON bandwidth is already giving way to G-PON with the need for more bandwidth being one of the key drivers. The question is whether even G-PON will be adequate beyond the next few years.
2. Developments in networking and optical technologies since the late 1980s have dramatically altered the landscape. In the late 1980s, Ethernet was considered suitable only for data traffic in the LAN. SONET and ATM were emerging technologies that were expected to dominate transport and switching for all services. Instead Ethernet has expanded beyond the LAN; with new Carrier Class features, it is displacing ATM and SONET in the MAN and WAN as the dominant new multi-service switching and transport technology. Fiber-optic technologies have also matured, and the costs of optical fiber and optical components have dropped dramatically. These networking and optical technology developments position switched Ethernet solutions to bring a level of performance and scalability to the access network that is unimaginable with TDM-PONs.

Switched Ethernet solutions have frequently been positioned as “active Ethernet.” In reality, at the current cost of fiber and optical components, switched Ethernet may be deployed in both passive and active outside plant (OSP) configurations at total systems costs (including electronics and OSP) that are competitive with TDM-PONs. The optimum choice between passive, active, or a combination of passive and active switched Ethernet options depends on various factors such as the level of aggregation, subscriber density, feeder fiber availability, and other factors. The choice may be a matter of preference for a variety of other reasons, too.

Switched Ethernet also offers hidden cost benefits over TDM-PONs that are yet to be recognized widely. One aspect not obvious is that construction techniques for the optical fiber plant can be simplified to lower cable installation costs. This is an important consideration, because construction is the most significant cost item for FTTX.

Another example is that carrier operating costs will be higher for TDM-PONs, because a high level of specialization for network design, installation, and maintenance is required. Ethernet, by contrast, is widely deployed because of its plug and play characteristics, it is a well understood technology, and trained personnel are readily available.

Figure 1. Switched Ethernet provides un-matched Bandwidth Scaling Future-Proofing the Network

Switched Ethernet provides performance and scalability not possible with TDM-PONs

By reviewing developments in telecom technology, the cost of fiber and transceivers today, and trends in optical component technology, it becomes apparent that the criteria that originally motivated TDM-PONs have largely lost their relevance and that TDM-PONs are no longer the best investment for the future.

From TDM to Packet Networking

In the late 1980s, the network was dominated by TDM switching and transmission. ATM switching and SONET were new, and the Internet was in its infancy. Ethernet was a LAN technology, and ATM was supposed to displace Ethernet all the way to the desktop.

Today we have a networking landscape that is quite different from that vision. Rather than ATM to the desktop, the reverse has happened: Ethernet held fast to its reign in the LAN, became the standard Layer 2 LAN switching technology, and has expanded aggressively into the metro and onto the WAN.

Furthermore, the first deployments of Ethernet in the metro, based on simply extending enterprise Ethernet, are rapidly being displaced by “Carrier Class” Ethernet, a.k.a. Carrier Ethernet. Carrier Ethernet embodies five key attributes not found in enterprise-class Ethernet:

1. “Hard” quality of service (QoS) capable of guaranteed bandwidth and latency
2. Resiliency capable of 50 msec restoration in mesh and ring topologies
3. Scalability in several dimensions, including bandwidth, number of users supported, services, and services per user
4. TDM support that enables Ethernet to carry legacy TDM services
5. Operations, administration and maintenance (OAM) with network and service management features that exceed the capabilities of legacy TDM networks

These features have been integrated into Ethernet while preserving the ease of deployment and plug-and-play characteristics that are inherent to Ethernet.

Combining the powerful capabilities of Carrier Ethernet with Ethernet's ease of deployment gives it capabilities that cannot be matched by TDM-PON in CapEx, OpEx, services supported, flexibility, or scalability.

Basic TDM-PON Premise

B-PON, E-PON, and G-PON are close derivatives of the original TDM-PON concept. The basic premise of these PON architectures is to share the optical feeder and a port on the central distribution unit, known as the optical line terminal (OLT), among up to 32 or 64 subscriber access terminals. The subscriber terminals are called optical network termination (ONT) equipment. Passive optical splitters terminate the optical feeder and provide optical connections to the ONT.

The primary purpose of a TDM-PON protocol is to function as a traffic manager that manages bits on the point- to multi-point, shared-bandwidth optical connection created by the splitter-based architecture. It must abstract point-to-point connections on the PON using virtual circuits while maintaining security in a broadcast environment. This adds significant processing complexity and creates many compromises--most critically, an inability to scale.

For example, if the average committed information rate (CIR) to each customer increased to 100 Mb/s in a 1x32 TDM-PON, the system would have to operate at 3.2 Gb/s, and in a 1X64 TDM-PON, the system would have to operate at 6.4 Gb/s. If in the future the average CIR became 500 Mb/s, the TDM-PON would have to operate at 16 Gb/s. And so on. These are extremely high speeds and they can be unrealistic operating speeds for a TDM-PON from technical and cost perspectives.

Upgrading TDM-PONs is also very complex. High bandwidth demand by a few customers would force an upgrade of a whole PON including the OLT and every attached CPE. Alternatively, the split ratio on a TDM-PON can be reduced by allocating more bandwidth per customer. Reducing split ratios are at best a temporary measure because, in the end, it would turn a TDM-PON system into a point-to-point fiber system (no splitters). At this point, the TDM-PON protocol becomes redundant and, in fact, a liability because of its complexity.

Switched Ethernet

By contrast, switched Ethernet systems are inherently simple. They operate on a point-to-point optical circuit over dedicated fibers or over shared fiber with a wave-division multiplexed (WDM) overlay. Because switched Ethernet systems have a much higher baseline performance (100 Mb/s at a minimum), no equipment changes would be necessary up to that bandwidth if the service provider's basic service were initially provisioned for a lower speed. Upgrading bandwidth to customers requiring more than 100 Mb/s is a switch port change at the central office (and possibly a CPE change) only for the customers requiring the higher speed connection.

Ethernet has repeatedly demonstrated cost advantages over every other networking technology, including ATM, SONET, and frame relay. This has been proven repeatedly over the years as Ethernet has consistently been enhanced to deliver new levels of bandwidth at record-breaking new cost points. This trend will repeat itself, and we are likely to see 1 Gb/s connection speeds becoming as commonplace as 10/100 Mb/s connections.

Fiber and Optical Costs Plummet

There were two key drivers that led to the TDM-PON architecture in the late 1980s. First, fiber and optical transceivers were very expensive, requiring an architecture that shared them where possible. Second, there was a desire to eliminate powered RTs, which at the time were mainly digital loop carrier (DLC) terminals and were considered high-maintenance items. The architecture also addressed the limited experience at the time with managing large numbers of fibers being terminated in COs.

Today, the cost of fiber and optoelectronics is a small percentage of what it was in the late 1980s. The incremental cost of fiber,* which was over $300 per mile, is today about $30 per mile. Connectors that were over $100 are now less than $15, and fiber splicers that were $30,000 are now $5,000. Laser transceivers that were thousands of dollars are now less than $50 in small form-factor pluggable (SFP) modules. These new cost points make it possible to build point-to-point fiber networks without intervening active RTs over much longer distances than previously possible.

* It should be kept in mind that trenching and installations costs are about the same regardless of the actual cable size. Therefore, for new deployments of feeder cable, it is the incremental cost of fiber that is the key factor that needs to be considered.

Figure 2. Cost of a 48-Fiber Cable Curve

Fiber and optical component costs are now a fraction of what they were when TDM-PON technology was first conceived.

Even in the U.S., where the longest loop lengths are encountered, a majority of residential customers who are within carrier serving area (CSA) distances can now be served with point-to-point fiber connections directly from the CO or existing RTs at price points that are comparable with TDM-PONs.

For distances exceeding CSA reach, or for situations where an existing cable with a limited number of available fibers can be leveraged as fiber feeders, active Ethernet RTs are often the most cost-effective solution, depending on customer concentration and aggregation levels. OSP components required for Active RTs have also improved significantly with mean time between failures (MTBFs) today measured in years . However, where the physical-layer characteristics of TDM-PONs are required or preferred, Ethernet over WDM (WDM-PON) is becoming an increasingly compelling solution, with the immense scaling capabilities of point-to-point switched Ethernet.

WDM-PON Emerging for the Future

In the mid 1990s, dense wave division multiplexing (DWDM) emerged as the killer solution for supporting long-distance fiber routes. In the early 2000s, DWDM also found widespread application in metro networks. These two developments spawned significant new innovations in passive and active fiber component technology, drove costs down, and helped manufacturing processes mature.

There have been innovations in passive thin film and waveguide-based splitters, couplers, and wavelength selection devices. There have also been extensive developments in active components that have enabled tight frequency control and optical amplification. These developments combined with maturing of manufacturing processes position coarse WDM (CWDM) and DWDM for economically realistic deployment in the access network. The key ingredient that needs to be added to bring down costs into the correct range is volume. This is what FTTX in general and fiber-to-the-premises (FTTP) in particular are set to do that long-distance and metro networks could not.

Summary

TDM-PONs were conceived in the late 1980s at a time when the expectations for bandwidth scaling were much lower, TDM technologies dominated the network, and optical technologies were primitive and expensive. Since then, Ethernet has emerged as a proven multi-service networking technology, with a demonstrated ability to deliver new levels of performance at record-breaking cost points. In the same time period, optical fiber and optical components have matured and dropped significantly in price, and WDM is emerging as a viable option for the access network.

TDM-PONs were a concept that addressed the issues and expectations of the late 80s, but they cannot scale to meet the needs of the future. Switched Ethernet in both passive and active configurations provides unmatched performance and scalability and is the optimal solution for maximizing revenue generation potential for the future.

Continue

• Part 1:
  Why Rethink TDM-PON?
• Part 2:
  Debunking 10 Myths