Rootconf
Rootconf

Rows, columns, and consequences

Speak at Rootconf’s Special Edition on Databases

All submissions

Krithika Subramanian

You Can’t Just Plug In a Vector Index!! - Making Vector Indexes Work in Relational Databases

Submitted Apr 27, 2026

Session type - select the format for your session: 30-minute talk – technical deep dive

Vector search is increasingly central to AI applications, yet it is often implemented as an external system disconnected from relational query processing. This talk examines the challenges of implementing vector search inside a mature relational database, using SQL Server as a case study. We show that the key difficulty lies not in ANN algorithms themselves, but in integrating similarity search with cost‑based optimization, predicates, joins, transactional updates, and SQL correctness guarantees. SQL Server addresses these challenges by modeling vector search as a plan pattern composed of relational operators, with streaming execution semantics and optimizer awareness. This enables efficient filtered vector search, robust behavior under imperfect selectivity estimation, and naturally enables hybrid retrieval combining semantic similarity with structured queries directly inside SQL.

Target Audience: Database engine developers, Distributed systems engineers working on data platforms

Krithika Subramanian - Principal Engineering Manager - Microsoft.

All submissions

Comments

{{ gettext('Login to leave a comment') }}

{{ gettext('Post a comment…') }}
{{ gettext('New comment') }}
{{ formTitle }}

{{ errorMsg }}

{{ gettext('No comments posted yet') }}

Hosted by

Rootconf
Rootconf
We care about site reliability, cloud costs, security and data privacy