2 comments

[ 3.3 ms ] story [ 13.2 ms ] thread
So, which library was sleeping for 1 second?
It is an internal cache implementation. The library writer _thought_ this codepath would never be triggered - or at least very rarely, but that was proven wrong on the new hardware.

In my uninformed opinion: 1. waiting for 1 second sounds like a hack 2. The application should be doing cache writes asynchronously as as not to affect overall server latencies