Slow binary search execution

[Aklakan]

docker run --rm -it  aksw/rpt "integrate" "SELECT * { SERVICE <x-binsearch:vfs:https://databus.dbpedia.org/ontologies/purl.obolibrary.org/obo--chebi--owl/2024.02.01-161433/obo--chebi--owl_tag=sorted_type=parsed.nt> { <http://purl.obolibrary.org/obo/CHEBI_23367> ?p ?o } }"

The times are 2,1 s for 5MB and 2,8 für 980 MB. So for 5MB it takes nearly as long as for 1GB, whereas one one suspect it to be much faster on such small data.

[Aklakan]

The 5MB file is this one: https://databus.dbpedia.org/ontologies/dbpedia.org/ontology--DEV/2023.11.27-081000/ontology--DEV_tag=sorted_type=parsed.nt

[Aklakan]

Tracking the reads during binary search on sorted n-triples files using a wrapped ByteChannel shows that the approach with hadoop's Bzip2Codec currently accesses ~5 times more data.