Journal of the Chinese Institute of Engineers, Vol. 29, No. 7, pp. 1149-1160 (2006) 1149 NODE ARCHITECTURES FOR AGGREGATION OF TRAFFIC FROM ACCESS NETWORKS (Invited) Yabin Ye*, Hagen Woesner, and Imrich Chlamtac ABSTRACT This article surveys and classifies recent prototypes of node architectures for dynamic traffic aggregation in all-optical access-core networks. These networks act as back- bones for optical line terminators (OLT) or electronic switches that serve optical ac- cess networks. For reasons of economy they are typically constructed as unidirec- tional rings or busses. Well-known optical time and wavelength division networks and hybrid versions following the photonic slot routing principle are introduced and compared. All of the architectures shown here perform some kind of carrier sensing to avoid collisions in a de-centralized way. We give approximations of the expect- able gain of dynamic channel sharing over static assignment and the tuning time penalty. Technological limitations and cost driving factors are analyzed with respect to a low-cost implementation of a dynamic traffic aggregation. Key Words: traffic aggregation, access-core network, dynamic sharing, bursty. *Corresponding author. (Email: yabin.ye@create-net.org) The authors are with the Create-Net, Via Solteri 38, Trento, 38100, Italy I. INTRODUCTION In recent years FTTx (fiber to the curb/build- ing/home) solutions have become economically feasible. Today the bandwidth provided by these technologies can be 50 or 100 Mbit/s per user. Among the possible technologies in access networks, the most promising one is the Passive Optical Networks (PON), which shows decreased operational expenditures (OPEX) and reduced capital expenditures (CAPEX) due to reduced port numbers that arise from multi- plexing the traffic from multiple Optical Networking Units (ONUs) onto a single fibre served by an opti- cal line terminator (OLT). Depending on the type of PON (A/B/E/GPON), the total bit rates today can be between 155 Mbit/s and 2.5 Gbit/s. Next generation PONs are being developed (named SuperPON, Talli and Townsend, 2005) which will lead to uplink band- widths of 10 or even 40 Gbit/s. At the same time the development of wavelength division multiplexing (WDM) and optical time divi- sion multiplexing (OTDM) transmission technology allows for hundreds of wavelengths in the core network with bit rates per wavelength of 640 Gbit/s and even higher (Yamamoto et al., 1998). The trans- mission capacity of one fiber can be several Tbit/s (Bigo et al ., 2000 and 2001, Farbert et al ., 2000; Ito et al., 2000). Nevertheless we still observe a gap between ac- cess and core networks. This has been called “the Metro gap” in recent years, but as the term “ metro network” has never really been defined people tend to understand different things and in fact one can find so-called “metro-core” and “metro-access” networks. In this sense, the term access core network is equiva- lent with the latter, lower part of metro networks. The so-called access core thus describes a second level in the hierarchy of networks that aggregates traffic from the access and delivers it into the metro or core networks, and vice versa. The observed “gap” is still there, but it is more of an economic nature, in that the technology has to become mature to make large- scale deployment affordable for network providers. To allow for this, there should be the following characteristics: 1) Dynamic channel sharing. One of the key features of Internet traffic is its bursty nature (Clark, 1999 and 2000). The peak rates from a user can be 100 times higher than the average rate. In current PONs, the number of ONUs per OLT varies between 16