Computer systems supported by photonic interconnects and photonic memory devices can reach performance and energy efficiency levels unattainable through purely electronic means across scales, from processor chips to the data center. However, the promised benefits cannot be realized through a simple replacement process; to reach their full potential, several aspects across the entire system stack may need to be redesigned. This thesis tries to explore opportunities to improve the energy consumption of existing data center level photonics as well as adopt emerging photonic devices to build fast and energy-saving cache memories. We identify and propose a number of specific architectural designs that aim at these goals. By studying their feasibility and effectiveness we have shown the positive impact of photonic devices on performance and energy efficiency. Our work also outlines the design and evaluation philosophies of integrating photonics more efficiently into existing computing systems, laying the groundwork for realizable designs that can be quickly adopted in real life.

Creator

• Han, Haiyang

DOI

• https://doi.org/10.21985/n2-s8m8-bc57

Subject

• Computer engineering

Language

• en

Alternate Identifier

• http://dissertations.umi.com/northwestern:16052
• etdadmin_upload_901541

Keyword

• Emerging optical and photonic technologies
• Data center network
• Cache hierarchy
• Nanophotonics
• Network-on-Chip
• Optical interconnects

Date created

• 2022-01-01

Resource type

• Dissertation

Rights statement

• In Copyright