Award Date
8-1-2013
Degree Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Electrical and Computer Engineering
First Committee Member
Mei Yang
Second Committee Member
Yingtao Jiang
Third Committee Member
Shahram Latifi
Fourth Committee Member
Amei Amei
Number of Pages
83
Abstract
The rapid advance of technology enables a large number of processing cores to be integrated into a single chip which is called a Chip Multiprocessor (CMP) or a Multiprocessor System-on-Chip (MPSoC) design. The on-chip interconnection network, which is the communication infrastructure for these processing cores, plays a central role in a many-core system. With the continuously increasing complexity of many-core systems, traditional metallic wired electronic networks-on-chip (NoC) became a bottleneck because of the unbearable latency in data transmission and extremely high energy consumption on chip. Optical networks-on-chip (ONoC) has been proposed as a promising alternative paradigm for electronic NoC with the benefits of optical signaling communication such as extremely high bandwidth, negligible latency, and low power consumption. This dissertation focus on the design of high-performance and scalable ONoC architectures and the contributions are highlighted as follow:
1. A micro-ring resonator (MRR)-based Generic Wavelength-routed Optical Router (GWOR) is proposed. A method for developing any sized GWOR is introduced. GWOR is a scalable non-blocking ONoC architecture with simple structure, low cost and high power efficiency compared to existing ONoC designs.
2. To expand the bandwidth and improve the fault tolerance of the GWOR, a redundant GWOR architecture is designed by cascading different type of GWORs into one network.
3. The redundant GWOR built with MRR-based comb switches is proposed. Comb switches can expand the bandwidth while keep the topology of GWOR unchanged by replacing the general MRRs with comb switches.
4. A butterfly fat tree (BFT)-based hybrid optoelectronic NoC (HONoC) architecture is developed in which GWORs are used for global communication and electronic routers are used for local communication. The proposed HONoC uses less numbers of electronic routers and links than its counterpart of electronic BFT-based NoC. It takes the advantages of GWOR in optical communication and BFT in non-uniform traffic communication and three-dimension (3D) implementation.
5. A cycle-accurate NoC simulator is developed to evaluate the performance of proposed HONoC architectures. It is a comprehensive platform that can simulate both electronic and optical NoCs. Different size HONoC architectures are evaluated in terms of throughput, latency and energy dissipation. Simulation results confirm that HONoC achieves good network performance with lower power consumption.
Keywords
Computer Architecture; Computers; Optical; High performance computing; Micro-ring Resonator; Network-on-Chip (NoC); Networks on a chip; Optical NoC; Optoelectronics
Disciplines
Computer and Systems Architecture | Computer Engineering | Electrical and Computer Engineering | Optics
File Format
Degree Grantor
University of Nevada, Las Vegas
Language
English
Repository Citation
Tan, Xianfang, "High-Performance, Scalable Optical Network-On-Chip Architectures" (2013). UNLV Theses, Dissertations, Professional Papers, and Capstones. 1955.
http://dx.doi.org/10.34917/4798032
Rights
IN COPYRIGHT. For more information about this rights statement, please visit http://rightsstatements.org/vocab/InC/1.0/
Included in
Computer and Systems Architecture Commons, Electrical and Computer Engineering Commons, Optics Commons