Tenon Stack

Work / Data platform · Performance

From thirty seconds to five.

AI analysis of slow-query logs on a 100-million-row database cut key query times from 30 seconds to 4–5, and reviewed every query pattern for injection risk.

The situation

A production database had grown past a hundred million rows, with individual tables above ten million. Screens that once loaded instantly were taking thirty seconds, and the slow-query log had become far too large for anyone to read through.

The team knew the database was slow. What they lacked was a way to tell which queries were responsible, which part of the application issued them, and whether any of them were unsafe as well as slow.

What we built

We built an analysis layer that reads the query logs in full. It groups similar queries, ranks the groups by the total time they cost, and identifies the area of the application each one belongs to, so a slow query is reported as a slow feature that someone can own.

The same pass reviews each query pattern for SQL-injection exposure and unsafe construction, turning a performance exercise into a security review at no extra effort.

What changed

Guided by the ranked findings, the worst queries were rewritten and indexed. Query times fell from around 30 seconds to 4–5 seconds, and every query pattern was reviewed for injection exposure in the same pass.

Database size
100M+ rows overall
Largest tables
10M+ rows each
Query time before
About 30 seconds
Query time after
4–5 seconds
Also delivered
SQL-injection risk review

The pattern behind it

Query Forensics: Slow-query logs read, traced to the part of the business they belong to, and fixed. Injection risks flagged on the way.

The workflow

From a noisy log to a ranked list of fixes.

  1. Query logsSlow-query and general logs from the database you already run.
  2. AI step: Group and rankSimilar queries are clustered and ordered by total time cost.
  3. AI step: Trace to the featureEach cluster is mapped to the screen, report or job that issues it.
  4. AI step: Review for riskPatterns open to SQL injection or unsafe input are flagged.
  5. Fix listDevelopers receive indexes, rewrites and patches in priority order.

Amber steps are the AI layer. Everything else is the system you already have.

Next step

Want the same result?

Tell us what your version of this problem looks like, and we will map the route to the same result.