Tag pprof

This is Part 2 of the pprof series. If you’re new to pprof, start with Part 1: Quick Start. Last week I had a service that was “slow.” The recommendation endpoint was taking around 28ms per request. Not terrible, but not great either. Management wanted it faster. I could have guessed. Maybe the database? Maybe JSON serialization? Maybe string operations? All reasonable guesses. All probably wrong. So I profiled it. The results surprised me.

I spent two weeks optimizing a function that accounted for 0.3% of my program’s runtime. Meanwhile, a JSON unmarshaling call that ran in a loop was eating 45% of CPU time. I just didn’t know it. This happens more often than you’d think. We optimize based on gut feeling, not data. We assume the database query is slow. We assume the algorithm is the problem. We’re usually wrong. pprof is Go’s profiling tool.