Re: [I] bthread signal & wait 问题咨询 (brpc)

[via GitHub]

GreateCode commented on issue #2849:
URL: https://github.com/apache/brpc/issues/2849#issuecomment-2768708248

   > > > > [@MJY-HUST](https://github.com/MJY-HUST) 
可以试试[#2907](https://github.com/apache/brpc/pull/2907) 看能不能解决问题？
   > > > 
   > > > 
   > > > 没效果
   > > 
   > > 
   > > [@GreateCode](https://github.com/GreateCode) 
[#2819](https://github.com/apache/brpc/pull/2819) 呢？
   > 
   > 这个改动较大，我摘出来测试下
   
   #2819 确实优化了性能，但从测试结果看，也存在调大pthread，同qps消耗cpu变大的情况。
   我感觉可能是bthread 
worker的wait/wake成本过大导致的，比如前者成本是5，bthread处理成本是10，有2个bthread在一个bthread 
worker处理消耗cpu 是20，但若被其他bthread worker steal走，则成本是10 + （10 + 5） = 25。
   所以最好是若bthread worker的bthread queue不长，就别让其他bthread worker steal了，可能bthread 
latency可能增大，但节省的cpu消耗是更期望。
   
   这是epoll优先处理的优化效果
   
![Image](https://github.com/user-attachments/assets/5fd80ca3-9d98-49e1-8f88-81b21fe6a645)
   
   这是在epoll优化基础上调整pthread的效果
   
   
![Image](https://github.com/user-attachments/assets/0153ebcc-9a45-4cd2-b985-93f6b1fc0cb1)
   
   
![Image](https://github.com/user-attachments/assets/ea6960f9-4e1b-4b16-b573-35d39995b0c0)


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: dev-unsubscr...@brpc.apache.org

For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@brpc.apache.org
For additional commands, e-mail: dev-h...@brpc.apache.org