Optimisation robuste de requêtes pour les SGBD parallèles multi-locataires // Robust Query Optimization for Parallel Multi-tenant DBMSs
|
ABG-139247
ADUM-75298 |
Thesis topic | |
| 2026-05-22 |
Université de Toulouse
Toulouse cedex 4 - Occitanie - France
Optimisation robuste de requêtes pour les SGBD parallèles multi-locataires // Robust Query Optimization for Parallel Multi-tenant DBMSs
- Computer science
SGBD parallèles multi-locataires, Optimisation de requêtes, Modèle de coûts, Robustesse
Parallel multi-tenant DBMS, Query optimization, Cost model, Robustness
Parallel multi-tenant DBMS, Query optimization, Cost model, Robustness
Topic description
Dans les environnements cloud, un Système de Gestion de Bases de Données SGBD installé sur une machine multi-processeur et partagé par plusieurs locataires est appelé SGBD parallèle multi-locataire. Le partage des ressources dans un tel SGBD permet aux locataires de payer pour les ressources qu'ils consomment, tout en maximisant la rentabilité du système. Pour cela, un contrat de niveau de service (SLA) est établi entre le fournisseur et un locataire. Ce contrat défini des niveaux d'objectifs de services (SLO) définissant la qualité de service à fournir au locataire. Pour atteindre ces objectifs, l'optimiseur de requêtes d'un SGBD engendre un plan d'exécution qui permet la satisfaction des SLO tout en considérant la rentabilité. Pour cela, l'optimiseur s'appuie sur des estimations calculées par un modèle de coûts. Ces estimations sont sujettes à des erreurs à cause de l'obsolescence de valeurs de paramètres, d'erreur dans le calcul ou de variation dans la charge de travail. Ces erreurs peuvent, d'une part, empêcher l'optimiseur d'atteindre ces objectifs et, d'autre part, engendrer une consommation excessive de ressources. L'objectif de cette thèse est de proposer des méthodes d'optimisation robuste de requêtes qui permettent à l'optimiseur de satisfaire les SLO en évitant une consommation abusive de ressources.
------------------------------------------------------------------------------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------------------------------------------------------------------------------
In cloud environments, a Database Management System (DBMS) installed on a multi-processor machine and shared by multiple tenants is called a parallel multi-tenant DBMS. Sharing resources in such a DBMS allows tenants to pay only for the resources they consume, while maximizing system profitability. A contract, called a Service Level Agreement (SLA), is therefore established between the service provider and the tenant. This contract defines Service Level Objectives (SLOs) that correspond to the quality of services provided by the provider to the tenant. To achieve these objectives, the DBMS query optimizer generates a query execution plan that guarantees SLOs satisfaction while considering system profitability. For this reason, the optimizer relies on estimations derived from a cost model. These estimations are subject to errors due to parameter value obsolescence, calculation errors, or workload variations. These errors could, on the one hand, prevent the optimizer from achieving its objectives and, on the other hand, lead to excessive resource consumption. The objective of this thesis is to propose robust query optimization methods that allow the optimizer to satisfy defined SLOs while avoiding excessive resource consumption.
------------------------------------------------------------------------------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Début de la thèse : 01/10/2026
------------------------------------------------------------------------------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------------------------------------------------------------------------------
In cloud environments, a Database Management System (DBMS) installed on a multi-processor machine and shared by multiple tenants is called a parallel multi-tenant DBMS. Sharing resources in such a DBMS allows tenants to pay only for the resources they consume, while maximizing system profitability. A contract, called a Service Level Agreement (SLA), is therefore established between the service provider and the tenant. This contract defines Service Level Objectives (SLOs) that correspond to the quality of services provided by the provider to the tenant. To achieve these objectives, the DBMS query optimizer generates a query execution plan that guarantees SLOs satisfaction while considering system profitability. For this reason, the optimizer relies on estimations derived from a cost model. These estimations are subject to errors due to parameter value obsolescence, calculation errors, or workload variations. These errors could, on the one hand, prevent the optimizer from achieving its objectives and, on the other hand, lead to excessive resource consumption. The objective of this thesis is to propose robust query optimization methods that allow the optimizer to satisfy defined SLOs while avoiding excessive resource consumption.
------------------------------------------------------------------------------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Début de la thèse : 01/10/2026
Funding category
Funding further details
Financement d'un établissement public Français
Presentation of host institution and host laboratory
Université de Toulouse
Institution awarding doctoral degree
Université de Toulouse
Graduate school
475 EDMITT - Ecole Doctorale Mathématiques, Informatique et Télécommunications de Toulouse
Candidate's profile
Distributed and Parallel Systems, Data Management Systems, Database Systems, Query Processing and Optimization, Cost Models, Cloud Systems, Programming Languages (e.g. C++, Java, Python).
Distributed and Parallel Systems, Data Management Systems, Database Systems, Query Processing and Optimization, Cost Models, Cloud Systems, Programming Languages (e.g. C++, Java, Python).
Distributed and Parallel Systems, Data Management Systems, Database Systems, Query Processing and Optimization, Cost Models, Cloud Systems, Programming Languages (e.g. C++, Java, Python).
2026-06-22
Apply
Close
Vous avez déjà un compte ?
Nouvel utilisateur ?
Get ABG’s monthly newsletters including news, job offers, grants & fellowships and a selection of relevant events…
Discover our members
ANRT
Généthon
Institut Sup'biotech de Paris
Medicen Paris Region
ONERA - The French Aerospace Lab
SUEZ
Ifremer
Aérocentre, Pôle d'excellence régional
Groupe AFNOR - Association française de normalisation
Servier
TotalEnergies
ADEME
Nokia Bell Labs France
Laboratoire National de Métrologie et d'Essais - LNE
Nantes Université
Tecknowmetrix
ASNR - Autorité de sûreté nucléaire et de radioprotection - Siège


