Characterizing opportunities for short reach optical interconnect adoption : a market survey and total cost of ownership model approach

Abstract

Over the past decade, the demand for digital information has increased dramatically with the rising use of the Internet and various types of multimedia data - text, audio, graphics, video, and voice. As a consequence, the technologies that connect and transport data have become critically important. Available interconnect technologies are broadly organized into two categories: electrical and optical. Although many digital systems use electrical interconnects, optical interconnects are becoming an attractive alternative as electrical connection has become increasingly difficult in terms of cost and performance. However, the transition from electrical to optical interconnects across multiple markets could still be hampered by its higher cost relative to interconnects in the mid-term. Thus, this work seeks to shed light on the following question: "What additional characteristics are useful to evaluate the attractiveness of optical interconnects in emerging markets?" This thesis seeks to explore and answer this question in three parts. The first part of the thesis attempts to gauge the opportunities and barriers to optical interconnect adoption in emerging markets through an analysis of first phase interviews with professionals working in the datacom, automobile, consumer hand-held device industries. Initial review of the response set shows that of the five initial emerging markets for optical interconnect, datacom, specifically high-performance computing (HPC), has the greatest potential for increased optical interconnect adoption in the near future. To further explore the environment for optical interconnects in the HPC, a second, more detailed questionnaire was distributed to a limited number of interviewees. In response to this interview, some respondents noted that several metrics other than cost and performance, particularly power consumption, as being "very important" when deciding which technology to adopt. The second part of the thesis is primarily concerned with investigating further the influence that power and performance concerns have on optical interconnect adoption in HPC data centers. Specifically, this part of the thesis seeks to explore whether power concerns in data centers could lead to increased adoption of optical interconnects. To that end, a cost model of an HPC data center has been developed to identify the possible economic impacts that the adoption of optical interconnect technologies would have in a power-driven scenario. The third part of this thesis presents a set of policy recommendations based on the results from the data center cost model.