September 27, 2017

Tobias Mayer: Self-optimizing BAR tolerant event dissemination


Publish/subscribe has emerged as a popular paradigm for building asynchronous distributed applications. It realizes a event-driven notification service that can be used as communication middleware. The main part of available contributions use client-server based architectures and some works propose more flexible peer-to-peer based systems. Cooperative distribute systems need to tolerate arbitrary (Byzantine-faults, BFT) as well as rational (selfish) behavior for correct operation. However, scientific works concentrate usually on robustness considering only BFT or rational behavior but not both. In recent years it has been payed increasing attention to the BAR model (byzantine, altrustic, rational). It considers BFT and rational behavior at the same time among cooperative (“normal”) behavior. There is currently a lack of BAR tolerant publish/subscribe architectures. This will be the first main contribution of the PhD by applying the BAR model to a given publish/subscribe system. The system will be considered as optimization problem to calculate an optimal-as-possible event dissemination structure. Then, a distributed BAR tolerant optimization technique – which is the second main contribution of the PhD – is used to solve this problem. This presentation outlines the PhD topic in more detail, the goals, contributions as well as the schedule.