When a video call stutters, an online game lags, or a webpage feels slow, internet speed is not always the problem. More often, the issue is network delays, known as latency. In his doctoral thesis, Simon Sundberg investigates how such delays arise and develops new methods for measuring them.
Most people associate network performance with bandwidth, that is, the amount of data that can be transferred per second. However, for many of today’s digital services, latency is just as important.
“If you are downloading a large film, bandwidth is usually the main factor determining how long it takes. But for interactive services such as video calls, online gaming and virtual reality, latency has a much greater impact on the user experience,” says Simon Sundberg.
In his thesis, What’s the Delay? Understanding Latency Across the Network, he examines latency across several parts of the internet infrastructure, including internet service provider networks, web servers and the satellite-based Starlink network.
The Largest Delays Occur Closest to Users
One of the main findings is that the largest delays often occur closest to the user, in the so-called “last mile” between the internet service provider and the connected device.
“We found that the greatest delays did not necessarily occur out on the wider internet. In many cases, the problems were located in the final stretch to the user, where traffic is often carried over Wi-Fi,” says Simon Sundberg.
The research shows that access networks remain a significant source of latency and that delays can sometimes reach one second or more. Such delays are clearly noticeable in applications such as video conferencing and online gaming.
Uncovering Hidden Delays
Improving the internet first requires the ability to identify and measure the underlying problems. For this reason, the thesis focuses on developing new tools capable of monitoring network latency in real time without disrupting traffic. These tools make use of technologies such as eBPF, which enables network traffic analysis directly within the operating system kernel with minimal overhead.
The measurements revealed that major latency spikes can occur within web server software itself, long before information is transmitted to the user. The research also identified how various mechanisms within Starlink’s satellite network influence network delays.
“Most of the time, the internet works extremely well, but occasional sudden delays can cause a video call to break up or an online game to freeze. Our research shows that these problems can have many different causes and may arise in several different parts of the network,” says Simon Sundberg.
Tools Already Being Used in Practice
The results of the thesis have not remained confined to academia. Several of the tools developed as part of the research are already being used by internet service providers and global Content Delivery Networks (CDNs). These tools make it possible to continuously monitor latency and identify user-affecting problems more quickly.
“By making latency visible, network operators and service providers gain a better understanding of where problems occur and how they can be addressed. Ultimately, the hope is that this will contribute to a better internet experience for everyone,” says Simon Sundberg.
Read the Thesis
