infrastructure solutions: Once APON A Time

Two Passive Optical Access Options Went Head to Head...

Both ATM passive optical networking and Ethernet PON have attracted attention for their promise to relieve the access network bottleneck, but there's a great deal of illusion and misunderstanding about the two options. APON and EPON can best be evaluated by considering their service range and associated costs.

Service range refers to the extent of network traffic transport capability for a diverse network applications (see chart).

Of course, service providers choosing a PON likely would have a specific function required in the area of service requirements. One such prerequisite can be that the PON guarantee quality of service for voice and video. The question to ask is, can the PON system generate the behavior that satisfies the function? Additionally, the behavior is generated by the structure. A PON system structure includes the architecture and the embedded protocols. Does the particular PON technology have the proper structure enabling the needed behavior? It is essential to go through these steps, starting from the service range, to ensure that the PON can provide the needed services. The cost then is weighed on the basis of the services. Only then can it be determined if there is a viable business case for the technology.

A Side-by-Side Comparison

Both APON and EPON share a point-to-multipoint architecture. The downstream traffic transport is broadcasting. The upstream traffic transport uses a TDMA protocol. Both architectures are packet-based access technologies. In fact, the APON community has pushed for synergy of the two at the physical layer of fiber optics.

APON, based on the ATM framework, is designed to be a cost-sensitive access solution. It has been standardized through the ITU-T 983.x recommendations. Commercial APON systems are available for deployment of voice, video and data services today. Additionally, current APON systems carry Ethernet traffic seamlessly. The highly developed, mature, and standardized suite of ATM traffic management techniques and off-the-shelf ATM components gives APON a solid technological base. In a nutshell, APON can guarantee QoS for diverse network applications cost-effectively and is thus attractive to major service providers such as ILECs.

Meanwhile, the IEEE 802.3ah task force is developing an EPON standard as part of the Ethernet In the First Mile (EFM) project. This is an attempt to enable Ethernet for the first time on the access network to become a ubiquitous standard. EPON standardization is a work in progress and although many technical issues remain to be resolved, the EPON solution is beginning to emerge. EPON expects to reap the benefit of a large Ethernet base and wide industry acceptance. Once there is an Ethernet core, EPON could potentially carry native Ethernet traffic straight through the access network.

It is critical to distinguish the standard Ethernet solutions, including the gigabit Ethernet intended for the metro area network from the EPON technology under development. Standard Ethernet is based on CSMA/CD protocol, which PCs and other data equipment use to interface to the network. Lacking QoS mechanisms, standard Ethernet solutions cannot ensure the low-latency, low-jitter transport required for real-time, delay-sensitive traffic. Most Ethernet solutions are implemented for "best-effort" Internet traffic, such as e-mail, Web and file transfer.

For Ethernet to be a ubiquitous standard, it must go beyond this current "best effort". Simply throwing bandwidth at the QoS problem is not a long-term solution. Consider what is required for EPON to achieve this goal. It must start with a mechanism to differentiate classes of services to a level comparable to that of ATM. This has been the focus of IEEE 802.1D and 802.1Q, allowing Ethernet frames to carry a priority information field. Ultimately, if the technology expects to provide full services, EPON would have to guarantee QoS for diverse network applications by establishing priority-based traffic control and providing traffic shaping, as necessary.

Is EPON going to cost less than APON? Assume for a moment that both provide full access services. At the physical and optical layers, EPON has adapted standards similar to those of APON. At the transmission convergence layer, EPON traffic transport must solve the same engineering problems presented earlier to APON. An example of this might be the segmentation and reassembly or SAR functionality necessitated by the non-native traffic requirement. All such architectural capability comes at a cost. When these technological hurdles are carefully studied, the costs of the competing PON technologies might be very similar. When we look at the services supported by the two, it could be quickly concluded that APON subsumes the service range of EPON. For a service provider wishing to choose a PON technology, its service requirements would lead to a distinctive bias and a pragmatic choice.

Finally, is EPON going to be more efficient than APON? This question, related to cost, must be brought into the context of the expected service range. Previous studies show that ATM is not inefficient when carrying IP packets for data services. A recent study highlights an EPON carrying T1 traffic, with the bandwidth efficiency at around 34 percent. Yet, EPON can have an edge if it is used only for data services with the "best effort." There is a great deal of optimization possible for any given technology once the exact context in which it will be used is well understood. When selecting a technology to serve a specific purpose, a pragmatic choice can be made but only when we go through the necessary teleological reasoning guided by our inherent "bias."

Click here to view Diagram

Zheng-Yang Liu is principal engineer at NEC Eluminant Technologies Inc. He can be reached at liuz@eluminant.com.

Share this article: Email, Slashdot, Digg, Del.icio.us, Yahoo!MyWeb, Windows Live Favorites, Furl
Add this article feed to: RSS, My Yahoo, Newsgator, Bloglines