Just spent about three weeks improving performance of an RDBMS-backed app. Identified slowest common ORM operations and hand-wrote SQL instead. 63% perf improvement. Got a proper DBA involved, added another 9 points.

app-DB average perv improved by 71%, est. client savings ~$85k/mo. 3 wks of programmer time and 1 wk of DBA time, and a few other things (scale testing etc. are expensive); billed $240k for the project. ROI in under 3 months.

But “machine time is cheaper than dev time”, right?