Dan Luu • 12/6/2021

Some latency measurement pitfalls

This article, based on a lightning talk from a Twitter engineer, explores historical pitfalls in measuring service latency. It highlights issues like opaque, uninstrumented latency in the network stack, lack of cluster-wide aggregation, and minutely resolution data. The author explains why server-side metrics often differ from client-observed latency and the challenges this creates for accurate monitoring and alerting.

0 comments

#distributed systems #observability #latency measurement