benchmarks of F2N on Raven testing the GTM MM access method

16 views
Skip to first unread message

George Lilly

unread,
Nov 4, 2011, 7:51:58 PM11/4/11
to fileman-tr...@googlegroups.com
*************************

benchmark of F2N qds load on raven with GT.M V5.4-002B Linux x86_64
and MM turned off

****************************


GTM>D IMPORT^C0XF2N("qds/qds.rdf")

STARTED: 3111104.152012
READING IN: qds/qds.rdf
73528 LINES READ
ADDED: _:G951670203 _S:805831840 fmts:rdfSource _TXT_INCOMING_RDF_FILE_/home/ge
o
rge/fmts/trunk/samples/qds/qds_rdf_394416761
71530 XML NODES PARSED
PARSE COMPLETE AT 3111104.152023
ELAPSED TIME: 8 SECONDS
APPROXIMATELY 8941 NODES PER SECOND
TRIPLES COMPLETE AT 3111104.153436
ELAPSED TIME: 853 SECONDS
APPROXIMATELY 81 TRIPLES PER SECOND <------ this calculation is
wrong, because "batches" of 10000

nodes at a time were processed mixed in with the dom node->triple processing
INSERTING 69537 TRIPLES
INSERTION COMPLETE AT 3111104.152023
ELAPSED TIME: 127 SECONDS
APPROXIMATELY 547 NODES PER SECOND <------ I only wish that this were
true, but alas it is based

on the time to process the last "batch" divided by the total number of triples
ENDED AT: 3111104.153643
ELAPSED TIME: 991 SECONDS
APPROXIMATELY 70 TRIPLES PER SECOND <----- this "overall" rate is the
one to watch...

*************************

benchmark of F2N qds load on raven with GT.M V5.4-002B Linux x86_64
and MM turned on

****************************


GTM>D IMPORT^C0XF2N("qds/qds.rdf")

STARTED: 3111104.175524
READING IN: qds/qds.rdf
73528 LINES READ
ADDED: _:G802962321 _S:387624835 fmts:rdfSource _TXT_INCOMING_RDF_FILE_/home/geo
rge/fmts/trunk/samples/qds/qds_rdf_405517303
71530 XML NODES PARSED
PARSE COMPLETE AT 3111104.175535
ELAPSED TIME: 8 SECONDS
APPROXIMATELY 8941 NODES PER SECOND
TRIPLES COMPLETE AT 3111104.180951
ELAPSED TIME: 856 SECONDS
APPROXIMATELY 81 TRIPLES PER SECOND
INSERTING 69537 TRIPLES
INSERTION COMPLETE AT 3111104.175535
ELAPSED TIME: 130 SECONDS
APPROXIMATELY 534 NODES PER SECOND
ENDED AT: 3111104.181201
ELAPSED TIME: 997 SECONDS
APPROXIMATELY 69 TRIPLES PER SECOND

K.S. Bhaskar

unread,
Nov 4, 2011, 9:35:04 PM11/4/11
to fileman-tr...@googlegroups.com
This suggests that the load is not limited by the database but
probably something else, like calculations.

Regards
-- Bhaskar

--
Windows does to computers what smoking does to humans

Reply all
Reply to author
Forward
0 new messages