axiqia commented on issue #4414: Performce: the latency and total run time increase when invoker_user_memory increase? URL: https://github.com/apache/incubator-openwhisk/issues/4414#issuecomment-479441536 @style95 Thanks for you reminder. And I did't collect these metrics. I guess that for 100 requests, 1Gb memory is too small, so the latency is high, 5Gb memory is too large, so the latency is still high, and 4Gb memory is more suitable. Bellow is my test shell command and result. ``` set -x set -e tmp=tmpfile function runload { /usr/local/bin/loadtest --insecure -n 100 -c 1 -k -m POST -H "authorization: Basic MjNiYzQ2YjEtNzFmNi00ZWQ1LThjNTQtODE2YWE0ZjhjNTAyOjEyM3pPM3haQ0xyTU42djJCS0sxZFhZRnBYbFBrY2NPRnFtMTJDZEFzTW# cat $tmp cat $tmp >> alllog } function repeat { for t in `seq 1 1 15` do echo "iterator " $t runload sleep 20 done } for mem in `seq 1 1 5` do• date >> alllog echo "memory"=$mem*1024 >> alllog wskdev -d invoker -t local -e invoker_user_memory="$[$mem*1024]m" -e skip_pull_runtimes=True sleep 20 repeat done• ```  I have a few questions. 1. Every time I updated `invoker_user_memroy` and redeployed OW, I will sleep 20. But the first loadtest with `invoker_user_memroy=3*1034m` and `invoker_user_memroy=3*1034m`, I got 503 error(Service Unavailable). 2. Regardless of the first iteration, with the second iteration as the warm operation, the latency of each subsequent iteration should be less than the second iteration. But the latency of the subsequent iteration is unstable, and sometimes even higher than the second iteration. do you have any idea? Best regards!
---------------------------------------------------------------- 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] With regards, Apache Git Services
