For the moment, this is just a dump of some BayesBenchmark results. Eventually, I'd like to fix up this page so others can easily add their benchmark resu\
lts for different setups and backends.

All numbers are the average for three runs of the benchmark. All times are in seconds.

Here is a test run, with 2000 msgs per bucket (1000 in the forget buckets).

System is SUSE Linux 9.2, 2.8 P4 (HyperThread) 512MB RAM w/ Perl v5.8.5

  • Berkeley DB - 4.3.x
  • MySQL - 5.0.10
  • PostgreSQL - 8.0.3 (with pg_autovacuum running every 60 seconds)

Everything ran local to the box

  • Run using Mail-SpamAssassin 3.1.0-RC1 release


Phase

SDBM.pm

DBM
(Berkeley DB/DB_File)

SQL.pm
MySQL MyISAM

MySQL.pm
MyISAM

MySQL.pm
InnoDB

PgSQL.pm
fsync=false

PgSQL.pm
fsync=true

1a

86.12

124.94

375.97

163.92

184.65

164.23

222.54

1b

87.68

142.01

352.76

159.77

170.67

154.03

211.93

2

530.39

1555.36

872.14

663.12

682.71

929.31

1149.92

3

18.38

49.47

9.18

9.34

9.78

6.79

8.69

4a

36.82

61.56

218.99

146.36

377.80

202.19

253.36

4b

37.14

60.23

201.49

146.36

381.26

196.09

242.49

5a

273.90

902.26

547.98

364.13

380.06

507.15

710.41

5b

193.95

443.86

284.70

228.32

233.19

316.11

432.08

total

1264.39

3339.70

2862.54

1881.10

2420.13

2475.91

3231.41

  • No labels