100 Gigabit Ethernet - Applications, Features, Challenges

100 gigabit Ethernet (100 GbE) is widely envisioned as the next-generation Ethernet, driven by continued capacity growth in provider networks, new emerging high-bandwidth applications, and an evolutionary trend from TDM to packet transport, particularly in the metro space. We will describe some key application areas where 100 GbE will probably be employed first. We further discuss possible implementations on the physical layer (PHY) and argue for the need of an efficient framer and against a medium access controller (MAC). Particularly for MANs and WANs, the most prominent candidate for the required framing will be based on the optical transport network (OTN, ITU-T G.872/G.709). 100 GbE will be an IEEE 802.3 standard and will hence only define the access control of and the transport on the physical sublayer. However, the challenges for realizing 100 GbE are related to Ethernet switching and not to Ethernet transport.

I. Introduction

Over the last decade a strong and continued growth of data (IP) traffic has been observed. Growth rates for data traffic vary from 30% to more than 100% per year, depending on the type of network. For example, traffic growth in private networks might differ from provider networks offering remote file storage services for enterprises or from networks of incumbent providers offering residential services with a mix of data and voice traffic from wireless or wireline phone services. However, it is widely accepted that data traffic will continue to grow over the next decade. The total traffic increase within that decade may vary somewhere between a factor of ten (26% Compound Annual Growth Rate, CAGR) and one thousand (100% CAGR). Predictions of required transport and switching capacities in the future are generally difficult to make for a variety of reasons. First of all, new high-bandwidth applications might emerge that have a significant impact on the annual growth rate and hence on the overall traffic growth over a period of five to ten years. Second, human behavior is unpredictable so that interests and needs years from now can not be foreseen today. Third, technological developments might quickly and significantly change the way content data are delivered to the end user and therefore also altering requirements and traffic demands in provider networks. As an example, we will discuss in the next section the developments in data storage and their impact on local content provisioning and backhaul capacities of networks. Nonetheless these challenges in predicting service provider networks five or ten years from now it is clear that more packet switching capacity is needed to handle the growth of IP-based traffic.