Optimizing AI server performance requires a systematic workflow that starts with precise bottleneck diagnosis acros