Back to search

Boston University

Developing instance-optimal data systems for modern applications

Abstract

dc:description.abstract

As data systems are deployed across increasingly diverse diverse applications, \textit{instance-optimal} data systems have emerged in an effort to build tools that are capable of being custom-tailored to the application at hand without extensive re-tuning or redeployment. A data system is considered \textit{instantiated} when it is deployed in a specific \textit{environment} to process a particular \textit{workload}. Instance-optimal data systems, therefore, leverage both workload characteristics and execution environments to achieve superior system performance. Workload characteristics encompass factors such as \textbf{access skew} and \textbf{matching skew}. Access skew occurs when certain portions of a dataset are queried significantly more frequently than others, while matching skew in join operations arises when some records in one table match disproportionately more records in another table. The execution environment encompasses both \textbf{hardware constraints} and \textbf{privacy requirements}. From a hardware perspective, instance-optimal systems must operate efficiently under memory limitations while maximizing the benefits of modern storage technologies, such as NVMe SSDs. Additionally, evolving data management requirements can also demand adaptable systems that balance performance with constraints such as data retention policies and efficient storage management. This thesis explores the impact of both workload and environment characteristics on overall system performance. We formalize multiple instance-optimization problems to exploit workload and environment characteristics across different components of the database system. Furthermore, we propose efficient algorithms that impose minimal CPU overhead on storage-based systems. Our evaluation demonstrates that instance-optimal algorithms can significantly outperform the state-of-the-art techniques, achieving up to a 2\times performance

Degree

thesis:*
Name thesis:degree_name
Doctor of Philosophy
Level thesis:degree_level
doctoral
Discipline thesis:degree_discipline
Computer Science
Grantor
Boston University
Year dc:date.issued
2025

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Zhu, Zichen
Advisor dc:contributor.advisor
  • Athanassoulis, Manos
ORCID dc:identifier.orcid

Subjects

dc:subject × 3

Rights

dc:rights
Statement dc:rights
  • Attribution-NonCommercial-NoDerivatives 4.0 International
Language dc:language.iso
en_US

Identifiers

dc:identifier.*
Handle dc:identifier.uri
https://hdl.handle.net/2144/53397

Chain of custody

source
Harvested from
Boston University
Base URL
open.bu.edu/oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Zhu, Zichen. Developing instance-optimal data systems for modern applications. doctoral thesis, Boston University, 2025. https://hdl.handle.net/2144/53397