[Bloat] PhD thesis with results related to buffering needs on variable-capacity links

[Bjørn Ivar Teigen]

Hi everyone,

I defended my PhD in December. I hope some of the results are interesting
to the bufferbloat community.

The title is "Opportunities and Limitations in Network Quality
Optimization: Quality Attenuation Models of WiFi Network Variability"
Full text here: https://www.duo.uio.no/handle/10852/98385

Here is the abstract:

The topic of this thesis is the performance of computer networks as
measured by latency and packet loss.

The maximum capacity, also known as bandwidth, of computer networks has
been steadily improving for decades. As more capacity becomes available,
new and higher bandwidth applications emerge, further increasing demand for
network resources. Performance in terms of latency and packet loss has not
seen the same regular improvement. Many applications convey real-time
interactions, and these depend on low latency and low packet loss for
optimal performance.

Optimizing network latency is a complex task. Physical transmission time,
scheduling, queuing, and interactions between traffic sources can all
contribute to the latency we observe. The challenge of reducing latency in
the Internet has received increasing attention in recent years. This
attention has led to successful solutions to several issues causing
latency. Other problems remain open despite these efforts.

This thesis aims to extend our understanding of WiFi networks through the
lens of latency and packet loss performance. We approach this using the
quality attenuation metric. Quality attenuation is a way of combining
latency and packet loss into a single variable. A quality attenuation value
consists of a latency distribution combined with the probability of packet
loss.

The thesis begins by investigating which performance issues are most
prevalent in today's WiFi networks. We show that both queuing latency and
the WiFi protocol specification itself are significant contributors. By
building a model of the WiFi protocol behavior we quantify the performance
of the protocol in terms of quality attenuation. We find that significant
performance variability is an inherent consequence of the protocol design.
Having quantified how variable WiFi links are, we explore the consequences
of this intrinsic variability for the performance of end-to-end congestion
control algorithms running over WiFi.

Because of the intrinsic variability of WiFi, achieving stable, low-latency
networking with WiFi will require innovation on many levels of the network
stack. In particular, we prove that capacity-seeking traffic can not
achieve both high utilization and low quality attenuation over a typical
WiFi network unless WiFi networks become much more predictable than they
are today. We hope that the methods and results presented in this thesis
will hasten the arrival of the low-latency Internet.

Best regards,
-- 
Bjørn Ivar Teigen
Head of Research
+47 47335952 | bjorn at domos.no <name at domos.no> | www.domos.no
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.bufferbloat.net/pipermail/bloat/attachments/20230103/0f62cb1e/attachment.html>