1 Department of Photonics Engineering, Technical University of Denmark2 Networks Technology and Service Platforms, Department of Photonics Engineering, Technical University of Denmark
This thesis focuses on the challenges of scaling current network node technology to support connection speeds of 100Gbps and beyond. Out of the many exiting aspects of reaching this goal, the main scope of this thesis is to investigate packet processing (address lookup and scheduling), forward error correction and energy efficiency. Scheduling and address lookup are key functions and potential bottle necks in high speed network nodes, as the minimum packet/frame sizes in both the popular Ethernet protocol, as well as the Internet Protocol (IP) still remains constant (84B and 40B, respectively). Therefore, in order to support a single 100 Gigabit Ethernet link, the routing mechanism must be able to support address lookup and output scheduling of over 148 million packets per second (pps) leaving only a few nanoseconds for each packet. With the emerging standards for 400Gbps (400GE and OTU5) and discussions on how best to exceed the 1Tbps boundary, the packet processing rate requirements for future network nodes are likely to increase even further in the coming years. Hence, there lies a tremendous task in expanding and optimizing current technology and methodology to support these increasing requirements. Forward Error Correction (FEC) is already a standard component of the Optical Transport Network (OTN) protocol as a means of increasing the bitrate-length product of optical links. However, the requirements for higher bitrates also drive a requirement for higher spectral efficiency in order to squeeze more traffic onto the existing physical transmission systems. To do this, while keeping the bit error rate (BER) below acceptable levels, more advanced FEC schemes are required. This is a challenge: Not only do we need to increase the processing speed of the FEC to handle the higher throughputs. The more advanced schemes also require more complex calculations to process each bit. This thesis will investigate how both the standard OTN FEC as well as more advanced FEC schemes can be implemented for 100G and above operation. As the networks are expanded to run at increasingly higher speeds, an unfortunate by-product is higher energy consumption. While advances in the physical hardware production (e.g. better chip production techniques) somewhat reduces the problem, it is imperative to think energy efficiency into the systems from the early design stage to the actual implementation and operation. Similar to the now common practice within micro processors, recent research aims at dynamically balancing performance and energy consumption of optical communication systems based on the immediate capacity demands. This thesis will describe various ways of achieving this dynamic capacity/energy trade off, with special emphasis on the Celtic project “Elastic Optical Networks” (EO-Net) and on adaptive forward error correction in particular.