Hi,
I'm using GAE standard Python27, manual scaling with 5 instances, class B4 with bottle as my WSGI server.
Currently, only 2 of the instances receive requests, one of them at 40 QPS and one at 10 QPS.
90% of the requests are around 30-60 ms but the other 10% exceed 200 ms, it's critical for me to stay below 100ms.
I added logs to my handlers and even to bottle WSGI handler and in the traces, I see that sometimes it takes a request 100-300ms to get to the bottle handler.
Meaning, I lose 100-300ms even before bottle or my code is invoked.
How can I improve it so all my requests will be around 50ms?
Thanks!