Contributions To Building Reliable Distributed Systems

Sonia Ben Mokhtar 1
1 DRIM - Distribution, Recherche d'Information et Mobilité
LIRIS - Laboratoire d'InfoRmatique en Image et Systèmes d'information
Abstract : Thanks to the latest evolutions in hardware and networking technologies we live in a world where networked computing systems are everywhere ranging from small/medium daily objects (e.g., watches, smart phones, cars) to large infrastructures (e.g., cloud platforms and data centers). On top of these computing systems a plethora of software systems/applications are invading our daily lives. Because of their intrinsic distribution and the involvement of more and more parties with sometimes conflicting interests, these systems are becoming bigger and increasingly more complex and thus more subject to faults. In this manuscript we consider two types of faults : Byzantine faults and rational faults. Byzantine faults are the most generic type of faults caused by nodes (e.g., software or hardware components running in a physical machine participating in the system) that may behave arbitrarily (e.g., by crashing, being subject to a bug, being under the control of a malicious attacker). Rational faults are caused by nodes trying to maximize their own benefit without contributing their fair share to the system. Dealing with Byzantine and/or rational faults in large scale distributed systems has been and still is a very active field of research. This manuscript, which gathers a set of research works I conducted from 2010-2017 with my collaborators and PhD students, contributes to this field. It particularly presents the following contributions : — RBFT : a robust Byzantine-fault tolerant state machine replication protocol ; — FullReview : a robust Byzantine fault detection protocol ; — PAG : a privacy-preserving Byzantine fault detection protocol ; — FireSpam : a protocol for dealing with rational nodes in gossip-based spam filtering ; — Acting : a live streaming protocol that deals with colluding rational nodes ; — RAC : an anonymous communication protocol that deals with rational nodes ; — Seine : a framework for designing and injecting rational faults and — Racoon : a framework for automatically dealing with rational nodes. I finally sketch few perspectives in this very challenging and exciting research domain.
Complete list of metadatas

https://hal.archives-ouvertes.fr/tel-01936371
Contributor : Sonia Ben Mokhtar <>
Submitted on : Tuesday, November 27, 2018 - 1:40:07 PM
Last modification on : Thursday, January 31, 2019 - 10:23:42 PM

File

hdr.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : tel-01936371, version 1

Citation

Sonia Ben Mokhtar. Contributions To Building Reliable Distributed Systems. Performance [cs.PF]. INSA de Lyon, 2017. ⟨tel-01936371⟩

Share

Metrics

Record views

58

Files downloads

83