読み込み中...
It is not ok that we speed weeks, even months, trying to solve why software or hardware is slow. In the meantime, companies waste money on datacenter or cloud compute costs, users are unhappy with latency, and product evaluations can run out of time, leading to poor choices. It should not take more than a week to identify the root cause or causes for a system's poor performance, such that any issue reported on a Monday should be solved by Friday, or sooner. This talk explores a vision called "fast by Friday," a new approach to performance engineering made practical by eBPF superpowers, and explores engineering work we all need to do to make it possible.