Skip to Main content Skip to Navigation
Theses

A flexible and decentralised approach to query processing for geo-distributed data systems

Abstract : This thesis studies the design of query processing systems, across a diversity of geo-distributed settings. Optimising performance metrics such as response time, freshness, or operational cost involves design decisions, such as what derived state (e.g., indexes, materialised views, or caches) to maintain, and how to distribute and where to place the corresponding computation and state. These metrics are often in tension, and the trade-offs depend on the specific application and/or environment. This requires the ability to adapt the query engine's topology and architecture, and the placement of its components. This thesis makes the following contributions: - A flexible architecture for geo-distributed query engines, based on components connected in a bidirectional acyclic graph. - A common microservice abstraction and API for these components, the Query Processing Unit (QPU). A QPU encapsulates some primitive query processing task. Multiple QPU types exist, which can be instantiated and composed into complex graphs. - A model for constructing modular query engine architectures as a distributed topology of QPUs, enabling flexible design and trade-offs between performance metrics. - Proteus, a QPU-based framework for constructing and deploying query engines. - Representative deployments of Proteus and experimental evaluation thereof.
Complete list of metadata

https://hal.inria.fr/tel-03272208
Contributor : Abes Star :  Contact
Submitted on : Thursday, October 21, 2021 - 4:31:34 PM
Last modification on : Monday, December 6, 2021 - 5:12:03 PM

File

VASILAS_Dimitrios_2021.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-03272208, version 2

Citation

Dimitrios Vasilas. A flexible and decentralised approach to query processing for geo-distributed data systems. Computer science. Sorbonne Université, 2021. English. ⟨NNT : 2021SORUS132⟩. ⟨tel-03272208v2⟩

Share

Metrics

Record views

58

Files downloads

67