Question about using Qlever in an ETL pipeline

[C-Loftus]

Hi there thank you for your excellent work on Qlever.

Question

I was wondering if QLever, either now or in the future, is an appropriate and performant choice in an ETL scenario where there are many clients doing high throughput concurrent INSERT / DROP operations.

Context

• I am working on a program which syncs JSON-LD in an S3 bucket to a triplestore
• I find that Qlever works well for querying the final read only graph and wanted to explore using it for the staging / ETL graph as well

Optimizations?

• I have some flexibility and could also just start the graph fresh every time instead of calling DROP.
  • I wasn't sure if DROP has internal locks that would slow down other operations significantly
• I wasn't sure as well if there are any bulk load tips beyond just posting large sparql payloads over http like the following (but bigger obviously)

INSERT DATA { 
GRAPH <urn://example.com/graph> { 
                    <urn://example.com/subject> <urn://example.com/predicate> <urn://example.com/object> . 
}};

[C-Loftus]

From discussion here #2130 it seems like Qlever may be best suited at this time for read only workflows and anything that requires lots of concurrent updates may be best done in a different datastore. Thus closing this but would be happy to discuss with anyone if this is a similar pattern to anyone else's workflow