Profiling Performance Bottlenecks in Production (opens in new tab)
We used to have a background job that would take minutes to finish. Not "minutes" as in "go grab a coffee" — minutes as in: the user would kick something off, walk away, and start wondering if it was broken before finally sending a message to engineering. The job processed large sets of data, and the bigger the input got, the longer it took to finish. We cut this job’s runtime by roughly 80% with a change that, in hindsight, was almost embarrassingly small, but would have been hard to find by...
Read the original article