TL;DR stateless distributed compute across network doesn’t work well for highly stateful distributed HPC applications (duh).

However the point about “bringing data where code is” vs. “taking code where data reside”; isn’t this pretty much what Datomic Ions does? You deploy code to same nodes where the data are

I wonder how Ions would scale for those applications. :thinking_face:

thanks for the tl;dr, didn’t yet read it

but yeah, data+compute locality is a good thing I think, avoids shuffling of data. Lambda is more a stateless thing, connecting stuff together, so no locality there.

haven’t used Datomic in anger either. I’m more worried about database max datom count (maybe one could shard to different databases?) and write throughput (like batch ingestion, or IoT things), but maybe there isn’t a database that suits for all situations 🙂 I think Datomic could well suit for most business need

anyway, on Ions and query groups, nodes with JVM with cached data on just the kind of processing the group is supposed to do with the data, sounds just so right

I hope I some day get to work on this warm fuzzy feeling too 🙂