On 01/05/15 19:34, Mark D Wood wrote:
What is Fuseki's transactional model? I understand that TDB uses
transactions<https://jena.apache.org/documentation/tdb/tdb_transactions.html>,
but its bulk loader does not.
In particular, if I use the graph store HTTP
protocol<http://www.w3.org/TR/2013/REC-sparql11-http-rdf-update-20130321/>
and POST an RDF document, is that operation performed as a
transaction? (What constitutes a bulk load from the perspective of
Fuseki?)
Thanks, -Mark
Hi Mark,
Each HTTP request is a transaction, that includes the Graph Store Protocol.
If the storage is TDB, it becomes a TDB transaction.
If it's in-memory, then it is done with a MRSW lock (no abort).
There is no specific bulk load for Fuseki. The TDB bulk loaders work by
not having transaction overheads and also by manipulating the low level
files directly.
Currently, that's a size limitation. Outside Apache Jena, there is a
different version of TDB and I've done 25m triples loads with a 1G heap.
(25m just being a test size, not an innate limit). Getting that ready
is resource-limited.
Andy
