[SERVER-65290] Run snapshot_reads benchmark with 32/64 tickets and a single FIFO queue Created: 06/Apr/22  Updated: 20/Apr/22  Resolved: 20/Apr/22

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

Type: Task Priority: Major - P3
Reporter: Daniel Gomez Ferro Assignee: Jordi Olivares Provencio
Resolution: Done Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Sprint: Execution Team 2022-04-18, Execution Team 2022-05-02
Participants:

 Description   

In SERVER-65156 we noticed snapshot_reads had some strange behaviour. In the high concurrency cases, snapshot_reads_load_throughput improved considerably when reducing the number of maximum tickets, which was in line with other (write) workloads such as ParallelInsert. However snapshot_reads_with_load regressed a lot.

This test runs writes in half the threads and snapshot reads in the other half. Our guess is that the reduced number of tickets improved our write throughput, and our improved write throughput could have affected negatively our read throughput, possibly due to WT-9074

One way to reduce the read threads starvation would be to have them share a single FIFO queue with writes.


Generated at Thu Feb 08 06:02:21 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.