Date of Award

Fall 2023

Project Type

Thesis

Program or Major

Computer Science

Degree Name

Master of Science

First Advisor

Radim Bartos

Second Advisor

Dongpeng Xu

Third Advisor

Stephen Wissow

Abstract

One-way latency is a non-trivial aspect of network quality that can reveal more about network performance than other quantities that are easier to measure. One-way latency is the amount of time from transmission of a packet along a network path to its reception. To measure one-way latency, transmission and reception times must be measured in the same time base. Thus the most important precondition for accurate one-way latency measurement is time synchronization between the transmitting and receiving nodes on the network.

Our research is designing a feasible method for measuring accurate one-way latencies using the Global Navigation Satellite System (GNSS). This GNSS-based method is able to provide precise measurement results regardless of network asymmetry by synchronizing the Source node and the Destination node to the same time base. We equipped several Raspberry Pi platforms with GNSS signal receivers. In our experiments, those platforms were set in multiple American and Chinese cities, including Durham, NH, Dover, NH, San Francisco, CA in the US, and Beijing in China. The Internet paths between these locations offered a diversity of regional, transcontinental, and transoceanic practical network paths. Our method was validated by multiple existing tools. Our long-term measurements have spanned over one year, validated by multiple existing tools, revealing the features of both commercial and academic networks.

The daily and weekly one-way latency measurement results of our research demonstrated that the domestic networks in the United States, even transcontinental paths between the East and West Coasts, are much more stable than the transoceanic networks between America and China. Our study also has shown that the daily increases in the transoceanic network's one-way latency tend to occur around Chinese midnight. This implies that the Chinese off-hour network loads impact the one-way latencies more than American demands on transoceanic network paths during the workday, and that the impact is most significant in the westward transpacific paths. In our weekly observations, the patterns of one-way latencies do not seem to change between weekdays and weekends. Additionally, significant global events, like the FIFA World Cup, did not affect the Quality of Service of transoceanic networks between China and the United States. However, the device updates and the underlying access network technologies do have a substantial impact on the latency performance.

Our accurate one-way latency measurement methods for a variety of real-world networks can be used for network protocol design and evaluation. The industry can also use our methods for practical network performance analysis under various network conditions. Additionally, learning the one-way latency of practical networks improves our ability to optimize end users' experience.

Available for download on Thursday, September 11, 2025

Share

COinS