Title
"Methods for signal transmission and signal quality estimation in large data centers"
Supervisors
Principal supervisor: Associate Professor Darko Zibar, DTU Fotonik
Co-supervisor: Professor Lars Dittmann, DTU Fotonik
Co-supervisor: Senior Principal Engineer Steen Bak Christensen, Mellanox Technologies
Evaluation Board
Associate Professor Anders Clausen, DTU Fotonik
Docent Magnus Karlsson, Chalmers University of Technology, Sweden
PhD Fotini Karinou, Microsoft Research Ltd., UK
Master of the Ceremony
Associate Professor Michael Galili, DTU Fotonik
Abstract
Digital information has become an incredibly valuable resource and is the driving source of many aspects of our current society. Because of that, there is a great demand for increasing the performance and capacity of the networks which we use to process and communicate digital information. A large part of the heavy computations required for Internet Services and other large-scale problems happen within dense, high-performance networks, hosted in spaces such as Data Centers and Supercomputers. It is in fact within these large-scale machines where most of the global Internet traffic happens nowadays. Thus, it is valuable to investigate how to improve the performance of these networks in order to support the demand for services and interconnection desired by global communities.
The research in this thesis covers a selection of topics that are of interest for the development of Data Center interconnects (DCIs) with increased capacity.
The contributions can be classified in three groups: first, an exploration of the use of advanced modulation formats in the context of Data Centers, with an emphasis in multi-dimensional signaling for short-reach applications. Second, the design and system-level performance evaluation of integrated optical systems in two forms of the Silicon Photonics platform. Third, the investigation of quality metrics for transmitter testing in new interconnects using multi-level modulation formats.
Although seemingly unrelated, these three topics reflect on relevant research needs of the DCI industry within different time frames. First, looking at long-term and large-scale technological transitions, then, evaluating upcoming manufacturing platforms, and lastly, identifying short-term modeling and testing solutions for supporting immediate design problems.
In summary, the results presented in this Industrial PhD thesis have served as a base to guide further applied research in next-generation optical interconnects.