Filipa Salema Roseta Pedrosa

LoCaPS: Localized Causal Publish-Subscribe

Tese submetida para provas de mestrado em Engenharia Informática e de Computadores Instituto Superior Técnico, Universidade de Lisboa.

Abstract

This thesis addresses the problem of offering low latency to subscribers in a reliable causal publish-subscribe system. The publish-subscribe abstraction has emerged as a fundamental tool to build distributed systems that preserve strong decoupling among information consumers and producers. The most common strategy to implement this abstraction in large-scale systems consists of using a network of message brokers that relay events from publishers to consumers. These brokers require coordination to offer quality of service guarantees to message subscribers. In most systems that enforce reliability guarantees, a subscriber needs to wait until its subscription has been propagated to every broker in the system, and known by all relevant publishers, before starting to receive events. Curiously, this happens even when a subscription is covered by a previously deployed one. To the best of our knowledge, previous reliable causal systems do not focus on reducing the observed latency by subscribers nor on the coverage relationships between subscriptions. In this thesis, we study the properties that need to be satisfied to reduce subscription latency. We also propose a new publish-subscribe system that leverages causal order multicast to offer low subscription latency when subscriptions achieve such properties. Experimental results show that our algorithm can outperform previous solutions in terms of subscription latency.

Publicações

LoCaPS: Localized Causal Publish-Subscribe
Filipa Salema Roseta Pedrosa
MSc Thesis. Instituto Superior Técnico, Universidade de Lisboa.
November, 2020.
Available BibTeX, MSC Thesis, and extended abstract, and mid-term report.
Reducing the Subscription Latency in Reliable Causal Publish-Subscribe Systems.
F. Pedrosa and L. Rodrigues
Proceedings of the The 36th ACM/SIGAPP Symposium On Applied Computing (SAC) Online, March, 2021.
Luís Rodrigues