研究目的
Investigating the provision of ultra-low-latency guaranteed-rate communications for Cloud Services over an Enhanced-Internet, supporting two traffic classes (Smooth and Best-Effort) with deterministic QoS guarantees and improved energy efficiency.
研究成果
The Enhanced-Internet architecture supports two traffic classes, enabling Smooth traffic flows to achieve ultra-low-latency, deterministic QoS guarantees, and improved energy efficiency over GR-VCS connections. The proposed scheduling algorithms and router designs significantly reduce buffer requirements and queueing delays, increase network capacity, and lower operational costs and greenhouse gas emissions.
研究不足
The study focuses on fixed-size packets and does not extensively address variable-size IP packets, which are common in current Internet traffic. The proposed solutions require changes to existing router hardware, such as the addition of an FPGA per linecard to manage schedules.
1:Experimental Design and Method Selection:
The study proposes an Enhanced-Internet architecture supporting two traffic classes with virtual-circuit-switched connections for Smooth traffic flows. A new router scheduling problem, the Bounded Normalized-Jitter integer-programming problem, is formulated and solved with a polynomial-time approximate solution.
2:Sample Selection and Data Sources:
The study involves simulations of a linear chain of 10 routers of size 4x4 operating with 40-Gb/s links, configured with scheduling frame size F=2000. Several hundred Smooth flows were routed end-to-end.
3:Several hundred Smooth flows were routed end-to-end.
List of Experimental Equipment and Materials:
3. List of Experimental Equipment and Materials: The design includes a Silicon-Photonics all-optical packet switch with minimal buffering. Routers use two periodic TDM-based schedules (Queue and Flow schedules) to support GR-VCS connections.
4:Experimental Procedures and Operational Workflow:
The study involves configuring routers with Queue and Flow schedules, updating these schedules when the router's traffic rate matrix changes, and simulating traffic flows to evaluate performance.
5:Data Analysis Methods:
Performance metrics include end-to-end delay, jitter, buffer sizes, and energy efficiency. The study compares the proposed scheduling algorithms (Static-GPS, Dynamic-GPS, Random) in terms of these metrics.
独家科研数据包,助您复现前沿成果,加速创新突破
获取完整内容