QUIC performance monitoring: implementation of Spin Bit in Chromium

The objective of this thesis is to implement an optional feature of the QUIC protocol - the Spin Bit - in the Chromium open-source browser, in order to make it possible to perform QUIC traffic measurements. QUIC is a network transport protocol, based on UDP, that avoids the use of traditional techniques to compute network measurements because its packets are encrypted. The Spin Bit is one of the few bits of the QUIC packet header that is left unencrypted, allowing network operators to deduce Round Trip Time of the QUIC connection based on the Spin Bit value. RFC 9000 describes the Spin Bit feature of the QUIC protocol as an optional feature and for this reason the majority of the state-of-the-art implementations of the QUIC protocol does not support it. In the first phase of this thesis work I implemented the Spin Bit algorithm in the Google Quiche library, that is the library of the Chromium browser which implements the QUIC protocol for both the Android and Desktop platforms. In a second moment, in the testing phase, I used in-path network observer tools (Spindump for Linux and TIMQuic for Android) to detect the implemented Spin Bit feature. The testing phase was divided into two main parts. The first part was aimed to detect the correct behavior of the Spin Bit in a controlled environment, where both client and server resided on the same machine.The second part consisted in analyzing the RTT values produced by Spindump and comparing the results with Ping measurements in a communication between Chromium browser acting as a client and server placed on AWS EC2 instances in London and Frankfurt. Results show that average Ping values of RTT compared to Spindump results are in the same range. More specifically, Spindump RTTs are generally slightly higher than Ping RTTs, since they are subject to application layer delay; at the same time, by removing some outliers from the analysis subject to spin bit limitations, Spindump RTTs estimate in a good way the latency perceived by the application. In conclusion, this thesis work has reported a first technique to measure QUIC traffic with Chromium and has paved the way for the implementation of more complex techniques, based on more than a single bit of the QUIC Header in order to provide more accurate network measurement estimations.