Hey everyone! We're actively investigating Medplum right now and I'm struggling a bit with trying to understand throughput for bulk insertion. I've got a synthetic dataset of full patient histories for ~4000 patients with ~1000 observations across multiple encoutners per patient. I'm trying everything I can to make importing this fast. Following along with https://www.medplum.com/docs/fhir-datastore/fhir-batch-requests I'm using batch mode as opposed to transaction mode and audit logs and audit events are turned off. Despite that its looking like its going to take around 70-90 minutes to insert this dataset on my very beefy dev machine. Is this a realistic expectation or is there anything I can do to improve this?

To clarify - this is on your localhost/laptop? For some benchmarks, we do have some data available here https://www.medplum.com/blog/medplum-performance-test-2026

Sorry I missed the notification for this but yes it was on localhost. I got the throughput to be a lot better via aggressively tuning Postgres for a bulk load situation as well as running many instances of Medplum locally beind a load balancer