ModBioSQL SemiBenchmarks

Important notes:

The systems, I would like to compare, are not really comparable (e.g. it seems I could tune PgSQL better than MySQL; I would expect the same speed for PgSQL as MySQL with InnoDB tables.).
Moreover, I do not think that the only important property of a system is the speed.
MySQLdb is a python driver for MySQL, while PyPgSQL, psycopg are drivers for PgSQL.
UniProt 2.0 := SwissProt 44.0 + TrEMBL 27.0 (for programmers).
BioSQL was populated with SwissProt 43.3 using BioPerl.
EMBOSS seqret was used to get 100,000 random sequences from indexed flat-file.

Parsing; UniProt 2.0
This is only measuring the parsing and generating INSERT queries with some consistency checking.
	SwissProt	TrEMBL
Size of the flat-file:	575M	2379M
Parsing and preparing for inserts:	10 min	52 min

Parsing and loading; UniProt 2.0
flat-file:	N/A ( 0 min)	N/A ( 0 min)
ModBioSQL; MySQLdb:	22 min	99 min
ModBioSQL; PyPgSQL:	22 min	92 min

Creating indexes; UniProt 2.0
flat-file:	27 min
ModBioSQL; MySQLdb:	46 min
ModBioSQL; PyPgSQL:	31 min

Creating constrains; UniProt 2.0
ModBioSQL; MySQLdb:	100 min
ModBioSQL; PyPgSQL:	68 min

Total time
BioSQL; SwissProt; MySQLdb:	7h 40 min
BioSQL; SwissProt; psycopg:	7h 23 min
ModBioSQL; SwissProt; MySQLdb:	48 min
ModBioSQL; SwissProt; psycopg:	41 min
ModBioSQL; UniProt; MySQLdb:	4h 24 min
ModBioSQL; UniProt; PyPgSQL:	3h 33 min

Size on disk (approx.)
flat-file; UniProt:	3,225M 766M 3,991M	dat files indexes total
BioSQL; SwissProt; +indexes; PgSQL:	1.12G
ModBioSQL; SwissProt; +indexes; PgSQL:	706M
ModBioSQL; UniProt; PgSQL:	3,041M 780M 3,812M	data indexes total

Reading (100,000) random sequences
flat-file:	1,111,1111 seqs/s
BioSQL; SwissProt; psycopg:	7,142 seqs/s
ModBioSQL; SwissProt; psycopg:	14,556 seqs/s
ModBioSQL; SwissProt; PyPgSQL:	3,600 seqs/s
ModBioSQL; UniProt; MySQLdb:	5321 seqs/s
ModBioSQL; UniProt; PyPgSQL:	2832 seqs/s