In the HSQLDB site I can see this:
"The database performance test package PolePosition was released in 2005 and showed HSQLDB 1.8.0 with persistent memory tables is an order of magnitude faster than traditional solutions. PolePosition updated their tests in 2011 and used HSQLDB 1.8.1 disk tables (lower performance) with sync-on-commit "to be fair to other databases". Even in the latest published test results, HSQLDB is much faster than db4o and JavaDB in the critical flat object tests (image above) and several other tests. Complex Java object access with SQL queries can be performed much faster than PolePosition tests suggest if the SQL queries are optimized for HSQLDB."
PolePos is open and they can take these optimized SQL queries if someone provides them. Is anybody up to the task? Would it be too much work? Source for the benchmark available in polepos.org
I think the latest 2.2.5 handles the tests that perfomed slowly much better, due to query optimisations performed by the engine. I might run the tests when 2.2.6 is finished and send feedback to the PolePosition project if necessary.
Sounds great, thanks
Looks good, but have one suggestion. Please put more info about the configs on page 1 to make it self contained. The website talks about innoDB but no mention in PDF itself. Also HSQLDB is IN_MEMORY, right?
Thought of linking on another forum, but too much to explain.
The PDF is created by the PolePosition software. Will add extra description of the settings.
HSQLDB is actually running with CACHED tables. As smaller data sets fit imostly n the memory cache, it improves the speed. However, all database engines cache some rows.