maskit commented on pull request #7719: URL: https://github.com/apache/trafficserver/pull/7719#issuecomment-827359999
I understand that batch delete would require double of memory space than current if we keep the same number of sessions in the cache, but would it be unacceptably large? I didn't put this because I thought we'll see similar results, but this is what I got for test 1, 2 and 3. <img width="373" alt="image" src="https://user-images.githubusercontent.com/153144/116195964-d9a9f480-a76d-11eb-928e-aeff3b1180fa.png";> I do think ATS shouldn't spend CPU time for session cache eviction too much, but a sad thing of this change is that we are completely losing entry prioritization. Prioritizing entries by creation time may be not a good strategy if the cache hit rate is as bad as random eviction. But prioritizing entries by other info (e.g. protocol version, algorithm, User-Agent, etc.) may work and it may reduce CPU usage in total if it improves cache hit rate, which could be done by just changing the key of `std::map`. I'd have to test this on prod if it changes the functionalities. -- 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. For queries about this service, please contact Infrastructure at: [email protected]
![https://user-images.githubusercontent.com/153144/116195964-d9a9f480-a76d-11eb-928e-aeff3b1180fa.png"](https://user-images.githubusercontent.com/153144/116195964-d9a9f480-a76d-11eb-928e-aeff3b1180fa.png"){kind=link}