Reads the optimizer's data
Execution plans, statistics, wait events and index usage, through a read-only account. It does not read table rows.
For IT heads and DBAs
Oracle · Microsoft SQL Server · MySQL
Find the queries behind a slow system, and the fix.
Reports run longer every quarter, the application times out at month end, and the usual answer is more hardware. Query Advisor connects to Oracle, SQL Server or MySQL through a read-only account and reads what the optimizer already knows: execution plans, statistics, wait events, index usage. It ranks the statements worth fixing and, for each one, Gemini on Vertex AI writes a suggestion a DBA can act on, with the reasoning and the trade-off spelled out.
Findings (14)
#12 Full scan on ORDERS
High (est.)#9 Missing statistics on INVOICE_LINES
Medium (est.)#7 Implicit conversion on CONTACTS.PHONE
Medium (est.)#5 Sort without index on SHIPMENTS
Low (est.)
Finding #12 · Full table scan on ORDERS
High impact (estimated)Seen 18,420 times in sample
SELECT o.ORDER_ID, o.ORDER_DATE, o.TOTAL FROM ORDERS o WHERE o.CUSTOMER_ID = :cust_id AND o.ORDER_DATE >= :from_date ORDER BY o.ORDER_DATE DESC;
Plan excerpt
TABLE ACCESS FULL ORDERS cost 48,210 (estimated) SORT ORDER BY
Add a composite index
CREATE INDEX IX_ORDERS_CUST_DATE ON ORDERS (CUSTOMER_ID, ORDER_DATE DESC);
Why: the predicate filters on CUSTOMER_ID, which is highly selective in the sampled statistics, and sorts by ORDER_DATE. One index on both columns serves the filter and the sort, so the optimizer can skip the full scan and the sort step.
Trade-off: one more index maintained on every insert into ORDERS. Plan cost after: 312 (estimated by the optimizer; verify in a test environment).
Export DDLMark reviewedDismiss
Suggestion by Gemini on Vertex AI · Requires DBA review
What it does
Ranks by estimated impact
Findings are ordered by how much work they cost the server, so your DBA starts with the ones that matter.
Writes the fix and the why
For each finding, Gemini on Vertex AI drafts the DDL or rewrite, the reasoning behind it, and the trade-off.
Covers the whole system
Indexes, query rewrites, statistics refreshes, partitioning and schema changes, across one workload.
Changes nothing on its own
Your DBA reviews each suggestion, exports the DDL and runs it in a change window. The tool has no write access.
Reaches on-site databases
An outbound-only connector in your network holds the credentials. Nothing opens inbound.
How a Query Advisor pilot works
A fixed-price Discovery first, then a fixed-price pilot on one workload. The three stages are described on the homepage; this is what they mean for Query Advisor.
-
You give
A read-only account on one database, or a plan and statistics export, and a DBA to review the findings.
-
We measure
Optimizer cost before and after on the suggestions your DBA runs, and elapsed time on the queries named in Discovery.
-
You get
A ranked list of findings for one workload, the DDL for each, and a report your DBA can take to the change board.
Pilot price fixed in writing at the end of Discovery.
Google Cloud services it uses
| Service | Used for | Why |
|---|---|---|
| Cloud Run | Collector API and analysis jobs | Scales to zero between collections. |
| BigQuery | Plans, statistics and workload history | Statistics are small and tabular; history stays cheap to store and query. |
| Gemini on Vertex AI | Reasoning and DDL for each finding | Model changes are tested against the pilot's evaluation set before they reach you. |
| Cloud SQL | Findings and review state | A transactional record of what was suggested, reviewed and dismissed. |
Most of the load is Gemini inference on Vertex AI for each finding and BigQuery for plan and statistics history. The deployment is its own Google Cloud project, in the region you choose.
On your side: a connector in a container or VM that holds the database credentials, makes outbound TLS connections only, and can be switched off by your IT team at any moment.
Security and data notes
Read-only account
The collector connects with an account that can read the catalogue and performance views. It has no access to table data and no write permission.
What leaves your systems
SQL text, execution plans, statistics and wait data, into your own Google Cloud project in the region you choose. Table rows stay in the database.
Logged and reviewable
Every collection and every suggestion is logged in your project. The audit trail shows what was read, what was suggested and who reviewed it.
Questions DBAs ask
Does Query Advisor read our table rows?
No. It connects through a read-only account and reads SQL text, execution plans, statistics and wait data from the system catalogue and performance views. Table rows stay in the database, and the connector makes outbound connections only.
Does our old Oracle or SQL Server version matter?
Usually not. Query Advisor reads plans and statistics through the system catalogue and performance views, which have been stable across versions for years. We confirm the exact version in Discovery and test the collector against it before the pilot starts.
Does it need an agent on the database server?
No. The connector runs in a container or VM anywhere in your network with outbound HTTPS and a route to the database. Nothing is installed on the database server itself.
What if our DBA disagrees with a suggestion?
Then it is dismissed, with the reason recorded. Suggestions are drafts for a DBA to judge; the tool ranks and explains, the DBA decides. Dismissed findings are kept so the same suggestion is not raised again without new evidence.
Talk to us about Query Advisor
Tell us which database, roughly how large the workload is, and what feels slow. The form is on the homepage, or write to contact@technometa.com.
Go to the contact form