Page Latches are used to guarantee consistency of index and data pages for user and system objects in-memory. Pages that manage allocations are protected by internal allocation pages (PFS, GAM, SGAM and IAM). This wait type accumulates while page latches are in update mode. Update mode is not just for UPDATE operations. It is used when SQL needs to read then modify a page, before modification it will place an update on the data then when ready upgrade this to an exclusive latch. If this wait type is high then the common problem of PFS contention may be present

Resolved by

DBAs and Developers

Suggested solutions

  1. Monitor page latches over time (see link below), confirm which database the problem is coming from. It is most likely tempdb
  2. Ensure tempdb has the correct number of data files. A rule of thumb is one file for every 4 CPU cores
  3. Enable trace flag 1118

Additional research

Analyzing Tempdb Contention
How to Tell if You Need More Tempdb Files
Tempdb Contention

⇐ Back to index