[SERVER-26953] Track bytes read into cache per query Created: 08/Nov/16  Updated: 15/Jan/19  Resolved: 15/Jan/19

Status: Closed
Project: Core Server
Component/s: Diagnostics, WiredTiger
Affects Version/s: None
Fix Version/s: None

Type: Improvement Priority: Major - P3
Reporter: Bruce Lucas (Inactive) Assignee: Bruce Lucas (Inactive)
Resolution: Duplicate Votes: 33
Labels: SWDI
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Issue Links:
Depends
depends on WT-2109 Add a method for retrieving informati... Closed
Duplicate
duplicates SERVER-38240 Extend OpDebug object to support stor... Closed
Related
Sprint: Storage Engines 2019-01-14
Participants:
Story Points: 0

 Description   

When slow queries are logged we include information about documents and index entries scanned to help diagnose performance issues. However that information doesn't tell us whether the documents or index entries scanned were in cache or not, and if they are not in cache they can have a much more significant performance impact. Reporting bytes read into cache for each query logged would help diagnose the cause of performance issues related to cache pressure.

This might be accomplished by tracking bytes read into cache per cursor or per session in WT, and computing the difference between the value of this counter before and after each query. Query performance impact would need to be evaluated.



 Comments   
Comment by Kelsey Schubert [ 15/Jan/19 ]

This work was completed under SERVER-38240 and is being documented under DOCS-12332.

Generated at Thu Feb 08 04:13:42 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.