Exponential Runtime for CREATE INDEX

We are working on a project that makes millions of insertions into the sqlite3 database, and faced a performance problem that is incompatible on different computers when called:

CREATE INDEX blocking_map_key_idx ON blocking_map (key)

A CentOS 5.8 runs this CREATE INDEX in less than 1 hour. Mac OSX Mountain Lion does not complete after 6 hours.

They have the same PRAGMA settings, the same version of Sqlite (3.3.6), and both seem to use the same amount of memory.

Is there a substantial way to execute sqlite on these two operating systems? If not, is it possible a disk or memory problem?

The full code that we run can be found here and is part of a larger open source deduplication library .

+5
source share

All Articles