Snowflake conceptsSnowPro CorePerformance
SnowPro Core performance: warehouse sizing and caching
What SnowPro Core asks about performance: the warehouse cache versus persisted results, result reuse, auto-suspend and query acceleration.
A persisted query result can be reused for up to 31 days after it was first run. Yet a query that differs only in upper or lower case does not get full reuse.
01 What does SnowPro Core test about sizing, caching and performance?
You rerun a query, it returns at once, and a question asks why. Before you answer, picture the two caches compared below.
Where this domain sits in the exam
The whole exam
- 1.0Snowflake AI Data Cloud Features and Architecture31%
- 2.0Account Management and Data Governance20%
- 3.0Data Loading, Unloading, and Connectivity18%
- 4.0Performance Optimization, Querying, and Transformation21%
- 5.0Data Collaboration10%
Save the 'Where this domain sits in the exam' graphic as a PNG image PNG for LinkedIn or Pinterest
A running warehouse caches table data for queries on that same warehouse, and drops the cache when it suspends. The bigger the warehouse, the bigger that cache. Persisted query results are a separate cache that expires after 24 hours.
The two caches: what each keeps and for how long
Warehouse cache
- Table data, for the same warehouse
- Larger on a larger warehouse
- Dropped on suspend
Persisted query results
- Results, for an exactly matching query
- Expires after 24 hours
- Reuse resets it, up to 31 days
The warehouse cache ends on suspend; persisted results last 24 hours per reuse.
02 When does a result get reused?
Next come the conditions. The new query must match the old one exactly, so a different case or a table alias blocks full reuse.
Functions such as RANDOM, external functions and hybrid tables also prevent reuse. Reuse is on by default and controlled with USE_CACHED_RESULT.
No exact match, no reused result.
03 How do sizing and suspension affect speed?
Then the warehouse itself. A larger warehouse has a larger cache, and shrinking a running one drops the cache of the removed resources.
AUTO_SUSPEND is set in seconds and defaults to 600, which is 10 minutes. Snowflake recommends at least 10 minutes for BI warehouses, about 5 for ad-hoc data science, and immediate suspension for tasks.
Suspend tasks at once, but give BI warehouses 10 minutes or more.
04 When does query acceleration help?
Some slow queries need neither. The query acceleration service offloads parts of outlier queries to shared compute, and suits large scans with an aggregation or selective filter. Queries with nondeterministic functions such as RANDOM are ineligible.
Before your next mock exam, write the two caches side by side from memory.
SYSTEM$ESTIMATE_QUERY_ACCELERATION tells you whether a query is eligible.
Summary
- The warehouse cache is dropped when the warehouse suspends.
- Persisted results expire after 24 hours; each reuse resets that, up to 31 days.
- AUTO_SUSPEND is in seconds and defaults to 600.
- Query acceleration suits large scans with an aggregation or selective filter.
Sources and editorial status
Research checked Nov 1, 2026.