|
The benchruns serve as an upper bound on the performance regressions when the PriorityTicketHolder is enabled. As each ticket is immediately acquired and then released, which is unlikely in a real system since a ticket usually is held for some amount of work.
|
|

With the new implementation of the PriorityTicketHolder SERVER-72072, the issue of "skipping the line" in the PriorityTicketHolder has been addressed. Above are the average performance numbers for each type of TicketHolder when ticketholder_bm.cpp is run 100 times.
|
|
With the completion of SERVER-72072, we are ready to revisit the benchmark performance of the PriorityTicketHolder
|
|
While performance is under investigation, we are working on SERVER-71829 to add latency distribution on the benchmarks so it is clear if the latency distribution is off in the future.
|
|
A full description can be found in the comments of SERVER-71213
|
Generated at Thu Feb 08 06:19:07 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.