This seems like a good conservative first step. I think it will be interesting to play with this value more in the future. I'm guessing we could go higher since most of the time in these requests is spent on I/O, not CPU. I've seen formulas like cores * 2 + 1 for similar generic parallel. It could also be that we want separate values per use of mapAsync depending on the use case.
Also, I came across this doc when researching this, which is pretty interesting:
This seems like a good conservative first step. I think it will be interesting to play with this value more in the future. I'm guessing we could go higher since most of the time in these requests is spent on I/O, not CPU. I've seen formulas like
cores * 2 + 1for similar generic parallel. It could also be that we want separate values per use ofmapAsyncdepending on the use case.Also, I came across this doc when researching this, which is pretty interesting: