MAHAMDI Célia
Team : DELYS
- Sorbonne Université - LIP6
Boîte courrier 169
Couloir 26-00, Étage 2, Bureau 234
4 place Jussieu
75252 PARIS CEDEX 05
FRANCE
Tel: +33 1 44 27 88 17, Celia.Mahamdi (at) nulllip6.fr
https://lip6.fr/Celia.Mahamdi
Supervision : Mesaac MAKPANGOU
Co-supervision : LEJEUNE Jonathan, SOPENA Julien
Distributed Multi-Consensus: Aggregation and Revocability
OMAHA (Opportunistic Message Aggregation for pHase-based Algorithms) is a message aggregation mechanism designed for phase-based algorithms, which are widely used in Cloud environments. In these settings, bandwidth is a critical resource, and managing small messages, which make up a large portion of network traffic, is a major challenge. OMAHA takes advantage of the predictable communication patterns of phase-based algorithms to group messages going to the same process. This reduces the total number of messages sent. Experiments show that OMAHA can save up to 30% of bandwidth while keeping latency increase below 5% when applied to the Paxos algorithm.
In distributed systems, reaching consensus on a decision or value is a complex problem, especially when processes have specific constraints. Systems like multi-agent setups (e.g., autonomous vehicles, schedule management, or robotics) and resource allocation systems must balance these constraints with a shared goal. Traditional consensus algorithms ignore such constraints, focusing only on proposed values. To handle faults, they rely on the majority, which means constraints from the minority are often disregarded. To address this, we proposed the f-Revocable Consensus. This approach ensures the chosen value respects the constraints of processes and allows decisions made by the majority to be revoked if they violate a minority process’s constraints.
Defence : 12/02/2024
Jury members :
Emmanuelle ANCEAUME, Directrice de recherche, IRISA [Rapporteur]
Gil UTARD, Professeur des Universités, MIS, Université de Picardie Jules Verne [Rapporteur]
Frédéric Le MOUËL, Professeur des Universités, INSA Lyon
Jonathan LEJEUNE, Maître de Conférences, Sorbonne Université, LIP6
Julien SOPENA, Maître de Conférences, Sorbonne Université, LIP6
Mesaac MAKPANGOU, Chargé de Recherche, Sorbonne Université, LIP6, Inria
2021-2024 Publications
-
2024
- C. Mahamdi : “Multi-consensus distribué : agrégation et révocabilité”, thesis, phd defence 12/02/2024, supervision Makpangou, Mesaac, co-supervision : Lejeune, Jonathan, Sopena, Julien (2024)
-
2023
- C. Mahamdi, J. Lejeune, J. Sopena, P. Sens, M. Makpangou : “OMAHA: Opportunistic Message Aggregation for pHase-based Algorithms”, 2023 IEEE 28th Pacific Rim International Symposium on Dependable Computing (PRDC), Singapour, Singapore, pp. 150-160, (IEEE Computer Society), (ISBN: 979-8-3503-5876-6) (2023)
-
2021
- C. Mahamdi, J. Lejeune, J. Sopena, P. Sens, M. Makpangou : “Réduire le coût de communication des algorithmes à phases par l’agrégation de messages : application à Paxos”, COMPAS 2021 - Conférence francophone d'informatique en Parallélisme, Architecture et Système, Lyon (virtuelle), France (2021)