Latches are used to guarantee consistency of non-buffer pool memory structures. This wait accumulates while SQL Server is waiting to process latch operations in keep mode. Meaning, the latch processed is collected while SQL Server is still determining the appropriate latch mode. Note that high values in this wait do not always mean a problem (e.g. in the case of parallelism)

Resolved by

DBAs and Developers

Suggested solutions

  1. If this wait is experienced further details will need to be retrieved from sys.dm_os_latch_stats
  2. If parallelism is a common latch wait consider optimizing heavy parallel queries via indexing

Additional research

Most common latch classes and what they mean
Diagnosing and Resolving Latch Contention on SQL Server

⇐ Back to index