has anyone ever benchmarked how quickly you can do a simple record.create(journal) in map.reduce?

its going to be wildly different in a multi-tenant setup. Easiest thing to do would just do your own log of the start/end and the difference. You could also use the APM bundle and run and get some metrics on it.

Ironically, I'm about to do an MR that creates 4-line ICJEs as part of its processing. NFT makes adding profiling info trivial so I can mention it here once I run it.

@Shawn Talbert would be very interested in seeing your results.

OK, FWIW, this is an AICJE with 4 lines on a slowish SB account: took 12.5 seconds to create and save. 😕

That’s longer than I hoped for - how many cores are assigned to your M/R?

if you mean 'processors' I'm running it as a single processor during development - though for a single JE case shown here, switching to parallel wouldn't make transactions save any faster. In fact, typically increasing parallelism for the same record type is more likely to slow the per-record save time?

Assuming a set (many thousands), I’d have hoped it would be the same speed per transaction but saving multiple at the same time.

that is true