Skip to main content

Server parameter tuning report

Overview​

At the end of every server parameter tuning session, DBtune generates a tuning report. It summarizes what the session did and what it achieved, comparing your database's starting configuration against the best configuration DBtune found.

The report answers the questions you care about after a session: did performance improve, by how much, which parameters changed, and which of those changes mattered most.

Where to find it​

The report is available in two places:

  • In the platform. Open your database, go to the Metrics tab, and select the tuning session under Tuning summary. The report appears once the session has finished.
  • By email. When a session completes, DBtune emails a PDF copy of the report (subject: "Your tuning session report is ready") to the relevant users, according to their notification preferences. The email is only sent when a session finishes normally, not when it is cancelled.

Baseline vs best configuration​

Every comparison in the report is between two configurations:

  • Baseline: the configuration your database started the session with.
  • Best: the best-performing configuration DBtune found during the session, which is the one automatically persisted at the end of the tuning session.

What's in the report​

Overview​

Headline numbers for the session:

  • Parameters tuned: how many parameters DBtune was allowed to change. See Parameter selection.
  • Number of configurations: how many distinct configurations DBtune explored during the tuning session.
  • Tuning time: how long the session ran.

Alongside these, the report shows the headline performance change between baseline and best:

  • Average query runtime for a standard tuning session, or
  • Fingerprint performance when the session targeted a Workload Fingerprint.

Additional metrics​

A baseline-vs-best comparison of the database and system metrics captured during the session. Depending on your database engine and hosting platform, this includes:

  • Transactions per second
  • CPU usage
  • Memory usage
  • Disk I/O (IOPS)
  • Temp files
  • Cache hit ratio
  • WAL generation

Metrics that aren't available on your platform are omitted.

Parameter changes​

The list of tuned parameters whose value changed, each showing its baseline value and its best value (with units). This is the configuration you can carry forward — the exact set of changes that produced the reported improvement.

Parameter importance​

This ranking identifies which parameters had the greatest impact on performance variability, helping you find out the parameters that actually drove performance shifts. Each parameter includes a confidence interval indicator representing the statistical reliability of its importance estimate.

To maintain focus, this list is a curated subset of your full parameter set. It prioritizes parameters that have:

  • Seen a significant performance change between parameter values.
  • Provided sufficient data to yield reliable statistical estimates.

Note on interpretation: The top-ranked parameters on the right do not necessarily represent the single factor that led to the most performance improvement in your best configuration. Instead, they identify the parameters with the highest overall influence on your workload's performance across all tested configurations.

Per-query comparison​

For Workload Fingerprint tuning sessions, the report breaks performance down per query. The queries that contribute most to total execution time are listed first, each with its average runtime and call count under the baseline and best configurations, so you can see exactly which queries the session sped up.

Last updated on