Concept and mechanism
A SARGable predicate lets the engine consider useful access paths but does not guarantee an index will be chosen. Applying functions to a column can make that access harder. To filter a year in a temporal column, a start-inclusive and end-exclusive range preserves every instant on the last day, using typed parameters. Statistics help the optimizer estimate cardinality; a large load or changed distribution can make estimates unrepresentative. Compare estimated and actual rows before forcing indexes. Also assess write maintenance and storage because improving one query can harm the batch supporting business close. Measure whole-workload results rather than optimizing a plan screenshot in isolation.
Guided application
In a fictional incident, DBCC CHECKDB reports corruption. Preserve evidence and assess sound recovery before using options that permit data loss. A physically consistent database after repair may still need functional reconciliation. For centralized monitoring, database watcher collects SQL-target data into an analytical data store and exposes dashboards. Inspected documentation still identifies it as preview. Check collector identity, target permissions, data-store access, and connectivity. Resource Reader for the portal viewer is not equivalent to collector-identity authorization. Define actionable alerts and validate sample age before declaring service health. Record collection gaps as uncertainty rather than interpreting them as evidence of no activity.
event_time >= @start_2026 AND event_time < @start_2027 preserves the interval without excluding December 31 times.
Common pitfalls
Cost as time; column function without measurement; repair as harmless; empty dashboard as health.
Related topics: Platform, capacity, and migration · Identity, networking, and encryption · Sensitive data and evidence
Measure execution and confirm integrity and collection independently.
Reference: Index design and SARGable predicates · DP-300 English objectives effective 2026-04-24