On Mon, 2013-05-20 at 18:36 +0300, Alexey Zakhlestin wrote:
> Is it possible to get the disk space used by the named graph in bytes?
> I didn't manage to find anything relevant by googling.
There's no way of doing that, because graphs are not stored apart from
each other. All RDF data reside in "shared" tables, namely
DB.DBA.RDF_QUAD for all quads (i.e., triples in graphs), DB.DBA.RDF_OBJ
for all distinct literals, DB.DBA.RDF_IRI for all distinct IRIs etc. If
a value appear in more than one triple it appears only once in the
"dictionary" table, so the footprint size should be "shared" between
"uses". Moreover, if all related triples are freed, the dictionary entry
remains, waiting for next reuse. Quads of DB.DBA.RDF_QUAD can be
counted, of course, but they're kept in heavily compressed form so the
actual disk size can not be calculated for a single graph --- each page
with data about that graph will probably contain data about other
graphs. To crown it all, there are remapped pages, they consume disk
Nobody cares about the disk, it's the cheapest component of a box
nowadays. A wonderful mechanics of 2TB sub-10-ms hard drive is cheaper
than the folded sheet metal of a PC case.