TY - GEN
T1 - Database is All You Need
T2 - 28th International Conference on Extending Database Technology, EDBT 2025
AU - Sun, Wenbo
AU - Li, Ziyu
AU - Srinidhi, Vaishnav
AU - Hai, Rihan
N1 - Green Open Access added to TU Delft Institutional Repository as part of the Taverne amendment. More information about this copyright law amendment can be found at https://www.openaccess.nl. Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.
PY - 2025
Y1 - 2025
N2 - Large language models (LLMs) have become central to many applications, but their deployment often requires high-performance hardware, specialized libraries, and complex engineering, limiting accessibility for smaller organizations. Meanwhile, relational database systems (RDBMS) are widely used for portability, efficiency, and native support for managing large-scale data operations. This paper presents TranSQL1, a toolkit that enables transformerbased LLM inference within RDBMS. By translating neural operations into SQL queries and representing model weights as relational tables, TranSQL leverages database features like dynamic disk-to-memory data management and caching to reduce hardware and engineering demands for serving LLMs. Using the LLaMA3 8B model, we demonstrate TranSQL's ability to implement attention layers, KV-cache, and end-to-end text generation through SQL queries. TranSQL offers a cost-effective, portable, and scalable approach to making advanced AI technologies more accessible.
AB - Large language models (LLMs) have become central to many applications, but their deployment often requires high-performance hardware, specialized libraries, and complex engineering, limiting accessibility for smaller organizations. Meanwhile, relational database systems (RDBMS) are widely used for portability, efficiency, and native support for managing large-scale data operations. This paper presents TranSQL1, a toolkit that enables transformerbased LLM inference within RDBMS. By translating neural operations into SQL queries and representing model weights as relational tables, TranSQL leverages database features like dynamic disk-to-memory data management and caching to reduce hardware and engineering demands for serving LLMs. Using the LLaMA3 8B model, we demonstrate TranSQL's ability to implement attention layers, KV-cache, and end-to-end text generation through SQL queries. TranSQL offers a cost-effective, portable, and scalable approach to making advanced AI technologies more accessible.
UR - https://www.scopus.com/pages/publications/105007908904
U2 - 10.48786/edbt.2025.103
DO - 10.48786/edbt.2025.103
M3 - Conference contribution
AN - SCOPUS:105007908904
T3 - Advances in Database Technology - EDBT
SP - 1118
EP - 1121
BT - Advances in Database Technology - EDBT
PB - OpenProceedings.org
Y2 - 25 March 2025 through 28 March 2025
ER -