A recent re-run of the classic Obama benchmark piqued my interest by returning a higher-than-expected latency. Taking the historic code and content used in T89775 for comparison, I'm getting the following results:
Current master:
git checkout 6aae433135a94f9dbf25915629986bc111fad558 rm -r node_modules && npm install node bin/server.js -n1 & ab -n 10 http://localhost:8000/en.wikipedia.org/v3/page/html/Barack_Obama/647485918 Document Length: 1779006 bytes Percentage of the requests served within a certain time (ms) 50% 15173 66% 15317 75% 15345 80% 15876 90% 16500 95% 16500 98% 16500 99% 16500 100% 16500 (longest request)
Old code:
git checkout 5d8b4e05c33e0533651800ae26d3c2e59bae90b9 rm -r node_modules && npm install node api/server.js -n1 & ab -n 10 http://localhost:8000/v2/en.wikipedia.org/html/Barack_Obama/647485918 Document Length: 2183602 bytes Percentage of the requests served within a certain time (ms) 50% 11865 66% 11933 75% 12170 80% 12660 90% 12751 95% 12751 98% 12751 99% 12751 100% 12751 (longest request)