Use Query Store to compare query plans and runtime performance across time windows, then investigate whether a slowdown follows a plan change, a workload shift, or resource contention. It preserves historical context beyond the plan cache, but it is an investigation tool—not an automatic explanation for every slow query.
What Query Store can—and cannot—tell you
Query Store records query, plan, and aggregated runtime-statistics history in time intervals. That makes it possible to compare periods even after the plan cache changes. It is available starting with SQL Server 2016, though features and management options vary by SQL Server version and platform. Microsoft describes its scope across SQL Server, Azure SQL Database, Fabric SQL database, Azure SQL Managed Instance, and Azure Synapse Analytics in its Query Store monitoring guide.
As an Amazon Associate I earn from qualifying purchases.
Query Store is not a per-execution trace. It stores estimated plans and aggregated runtime statistics; a recorded plan should not be mistaken for an actual plan captured for every execution. A query that has become slow may reflect a plan-choice regression, resource waits, contention, a changed workload, or other environmental changes. Query Store helps narrow the investigation, but does not prove the cause on its own.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Establish scope and confirm Query Store is collecting
First identify the environment: boxed SQL Server, Azure SQL Database, Azure SQL Managed Instance, Synapse dedicated SQL pool, or Fabric SQL database. Feature availability is not identical across these platforms and versions. Query Store begins with SQL Server 2016, and wait-stat tracking is available starting with SQL Server 2017 and Azure SQL Database according to Microsoft’s monitoring documentation.
Check the database’s Query Store state and configuration before interpreting missing or thin history. Configuration uses database-level ALTER DATABASE ... SET QUERY_STORE options. Microsoft’s monitoring guide covers administration and the available evidence.
Build a useful baseline before a slowdown
Compare stable, like-for-like periods—for example, normal business hours against the same hours on a prior comparable day. Record deployments, index or statistics maintenance, data growth, and workload changes when known; these events can affect plan choice or runtime independently of a software defect.
Track both broad resource consumption and the behavior of important user-facing statements. Query Store aggregates runtime statistics by interval, so it is useful for trends and comparisons, not reconstruction of every individual execution.
Rank #2
- Duration: useful for latency-focused questions; distinguish average from maximum and total duration.
- CPU and I/O: show compute and read activity, especially when assessing aggregate workload cost.
- Execution count: helps explain frequency, but a high count does not mean each execution is slow.
- Memory and waits: add context where supported and relevant to the workload.
Microsoft recommends a 900-second (15-minute) Query Store interval as a balance between capture performance and data availability; this is configuration guidance, not a universal performance optimum. See How Query Store collects data.
Find the query that changed
Use SSMS views for a first pass
In SQL Server Management Studio, inspect Query Store’s Regressed Queries view for recent degradation and Top Resource Consuming Queries for workload impact. Choose the time range and ranking measure deliberately. The documented views and some UI capabilities depend on SSMS and SQL Server versions; Microsoft’s best-practices page notes SQL Server Management Studio v18.0 and SQL Server 2017 or later for some views.
Ranking by total CPU answers a different question from ranking by average duration or execution count. A query run millions of times may dominate total resource use while having acceptable latency per run; a rarely run statement may have very high average latency but little total workload impact. Select the aggregation that matches the symptom.
Use catalog views when you need scripted analysis
For custom reporting, Microsoft’s documented Query Store catalog views include sys.query_store_query_text, sys.query_store_query, sys.query_store_plan, sys.query_store_runtime_stats, and sys.query_store_runtime_stats_interval. The official Query Store usage scenarios material includes examples for recent executions, execution counts, high physical reads, and queries with multiple plans. Adapt those examples to the correct interval boundaries and aggregation semantics rather than treating a single aggregate as a per-run measurement.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsDetermine whether the plan explains the regression
Compare plan history and metric trends over the same time window. Multiple plans or a decline that begins around a plan change are useful leads, not proof of causation. Microsoft’s Query Store usage scenarios calls a case where a new plan is significantly worse a “plan choice change regression.” The optimizer can choose another plan after changes in data cardinality, indexes, or statistics.
- Identify when the metric changed and which plan was active before and after.
- Compare the estimated plans and relevant runtime measures, such as duration, CPU, reads, or memory.
- Where supported, inspect Query Store wait categories for evidence of the kind of wait associated with the query or plan.
- Correlate the onset with a deployment, maintenance operation, parameter pattern, or workload shift only when the timing and other evidence support that connection.
If the symptom concerns a currently running request or an instance-wide incident, supplement Query Store’s history with live diagnostics. Microsoft’s monitoring guide also points to other SQL Server monitoring approaches, including DMVs and Extended Events; historical aggregates alone may not explain an active event.
Choose a proportionate response
Consider forcing a prior plan only when evidence supports it
If the query has multiple plans and the earlier plan performs better under the current workload, forcing it can be a targeted mitigation. SQL Server attempts to use the forced plan; if forcing fails, the optimizer can proceed normally. Recheck performance after applying it, monitor the workload for side effects, and unforce the plan when its justification no longer holds. Microsoft’s monitoring guide describes plan management through Query Store.
Understand the scope of automatic plan correction
Microsoft documents automatic plan correction for SQL Server 2017 and later: with Query Store enabled for workload tracking, the feature uses tuning recommendations to identify plan regressions and recommend a last-known-good plan. It is a supported option, not a universal replacement for diagnosis; validate suitability against the environment and workload. See Microsoft’s Automatic tuning documentation.
Recommended Free Tools
Keep Query Store data usable
Capture mode, retention, storage limits, and plan count affect how useful the history remains. Set them to fit the workload and the troubleshooting window you need, then monitor Query Store size and health. Microsoft’s best practices for managing Query Store discusses capture and management choices. For SQL Server 2016 users relying on just-in-time workload insight, Microsoft notes scalability fixes in KB 4340759.
When built-in monitoring is not enough
Query Store provides database-level plan and runtime history without requiring a separate commercial monitor. An organization with many servers may also need cross-server dashboards, centralized alerting, or estate-wide visibility; assess platform coverage, historical depth, incident visibility, operating burden, and current licensing terms before choosing another product. Redgate describes Redgate Monitor as offering multi-platform monitoring, query-performance analysis, alerting, and estate visibility, but those are vendor-stated capabilities; current commercial terms should be checked directly at its product page.
For readers learning how to interpret plans, Redgate lists SQL Server Execution Plans, 3rd Edition as a plan-focused book. It is optional learning material, not monitoring software; current availability can vary. See the publisher’s book page.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
