Afonso de Paiva e Pona Corte Real Gonçalves
Window Based Monitoring: Packet Drop Detection in the Network Data Plane
Tese submetida para provas de mestrado em Engenharia
Informática e de Computadores Instituto Superior
Técnico, Universidade de Lisboa.
Abstract
Detecting network anomalies, such as packet loss, is becoming an
increasingly important task to net- work operators, as applications
are becoming more and more performance sensitive. Several solutions
aim to detect these events as soon as they occur, as well as to
disclose where they are taking place. Unfortunately, some of them
incur in unacceptable overhead while others have to sacrifice coverage
in order to cope with the increasing traffic intensity. Software
Defined Networks and the Programmable Data Plane are relatively recent
technologies that allow network operators to configure how switches
process packets, which opens the door to efficient monitoring
solutions. In this thesis, we develop WBMon, a passive solution that
leverages the Data Plane programmability to perform an inter-switch
coordination algorithm that detects packet drops in arbitrary paths at
line speed. Additionally, we employ a Failure Inference Algorithm
(NetBouncer) to enable localizing the links responsible for packet
drops. Our evaluation shows that WBMon is able to detect every packet
drop in less than 2ms, which allows to detect short-lived failures.
Publicações
- Window Based Monitoring: Packet Drop Detection in the Network Data Plane
- Afonso de Paiva e Pona Corte Real Gonçalves
- MSc Thesis. Instituto Superior Técnico,
Universidade de Lisboa.
- November 2022.
- Available BibTeX, MSC Thesis, and extended abstract, and
mid-term
report.
Luís Rodrigues