Details
-
New Feature
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Fixed
-
1.4.1
-
None
Description
Vadim point out excessive spinlock usage as concurrency increase
https://www.percona.com/blog/2016/05/12/proxysql-versus-maxscale-for-oltp-ro-workloads/
An internal implementation discussion show up multiple set of task to get improvement :
- have N epoll_fd to remove one lock
- remove multiple X nonblocking polls and set it to 1
- implement thead pool