Impressions from the Measuring Network Quality for End-Users Workshop

The Internet has become a crucial part of users’ daily lives in the last two decades, with people using it to expand their social lives, to perform their daily jobs, to shop for a wide range of items, and to keep up with major events and news.

Even as the importance of the Internet has grown, some aspects of the end-user experience have not kept up. Many users have connection latency that remains at decade-old levels. Despite significant reliability improvements on some parts of the Internet, other end users continue to see interruptions in service.

To address these continuing problems with Internet service, the Internet Architecture Board hosted a workshop, called Measuring Network Quality for End-Users, on 14-16 September 2021. The IETF blog talked with Wes Hardaker, a member of the IAB and a senior computer scientist at the University of Southern California's Information Sciences Institute, about the goals of the workshop. An edited version of the email conversation follows.

An update from the workshop is part of the IAB Open session at 14:30-15:30UTC on 11 November during the IETF 112 meeting.

IETF Blog: What motivated the workshop?

Hardaker: The topic of end-user network quality has been discussed both in academic environments and in industry for a long time. The goal of the IAB workshop was to try to make forward progress and bring together segments of the industry that might be able to drive industry forward.

IETF Blog: Who participated in this year’s workshop?

Hardaker: We had 33 presentations from 45 papers. We had many segments of the industry, from academics to network service providers of various levels to operators attending. We don't have an exact count, but more than 90 people participated at some point during the event.

IETF Blog:  What were the biggest takeaways from the workshop?

Hardaker: The full list of conclusions can be seen in the initial workshop report draft. If I was to summarize the biggest conclusions they would probably be:

1. There are a large number of interconnected elements to both measuring and ensuring good network connectivity
2. Many layers of applications, network types and the Open Systems Interconnection (OSI) stack model are involved, exacerbating the problem.
3. The existing metrics of speed and latency are not sufficient alone, and too much of the time, measurements concentrate on  speed and not on actual Quality of Experience (QoE).
4. Significant work is needed to not just measure network quality, but also to use those measurements to stir the networking industry into action.

IETF Blog: What are the biggest contributors to latency? Are there any things that end users can do to limit latency?

Hardaker: Network bottlenecks and bufferbloat were heavily discussed in the workshop as major sources of latency issues. Many end-user networks in particular have devices that buffer significant amounts of data without any sort of prioritization that would allow high-priority/low-latency traffic to be prioritized.

As for what end-users can do, the general perception is that end-users are not directly in control of their network quality and typically don't understand the complex intertwining of things that cause their networking issues. As an example, one member experienced a significant degradation in his network performance when he microwaved a potato. The average end-user would have a difficult time diagnosing that problem, let alone fixing it. Thus, we must rely on service providers to figure out how to best measure the network on behalf of their users.

IETF blog: Were there any surprising issues brought up in the discussions that followed the presentations? If so, what were the most surprising?

Hardaker: I don't think anything was surprising, but a number of concise statements came out that nicely summarized networking issues in short succinct ways. Such as:

• Users don't need more bandwidth; they need better bandwidth.
• Measurements without a corresponding incentive to improve are functionally useless.

IETF blog: What might be some next steps, based on the workshop?

Hardaker: The workshop ended with a few thoughts about plans for future work. There was a general belief that there is both immediate work that might be proposed within existing IETF working groups, but yet, there are some longer-term issues that might be worth starting in an Internet Research Task Force (IRTF) research group.