Time/Place
- Time: 11:00AM Eastern Time US (UTC-5)
- Dial-in Number: (712) 775-7035
Attendees
Agenda
- Recent test results
- Virginia Tech Results - Test 2
- Virginia Tech Results - Test 3
- Virginia Tech Results - Test 5
- York University - Test 3 Results - LevelDB
- York University - Test 3 Results - MySQL
- Princeton Results - Test 5
- York University - Test 5 (In progress)
- Reporting out - enough tests have been run to create a summary picture for the community
- Action: Someone to draft a summary message to be added to the wiki and posted to the list
- Test 4; Complete
- Test 4 Results Summary
- https://groups.google.com/d/msg/fedora-tech/NoBhA0E14PA/bqHia6I6AwAJ
- Test 2
- Test 3
- Test 5
- Complete report
- Retrieving objects that link to a large number of repository objects (e.g., collections with thousands of members)
- Slow with LevelDB: 10 sec vs. 200 msec for non-repository URI objects
- Much slower with Postgres/MySQL: 30+ sec
- See Real World Performance (Retrieving Objects With Many Links to Repository Objects)
Minutes
- Additional test results have come in from Yinlin Chen. Nick Ruest is also running tests on an ongoing basis.
- In general, LevelDB performs well, but does not scale to large numbers of resources.
- PostgreSQL and MySQL scale much better.
- Database configuration and setup have a big impact on performance (unsurprisingly).
- We should write up summaries of each of the test, like the Test 4 summary
- Nick Ruest will write up summaries of Tests 2, 3, and 5
- Then we can move on to write up the overall summary
- Would be nice to get Test 1 results with large files, but need more storage to test effectively.
- Also good to run Test 6, which combines various operations.
- Esmé Cowles: I found that it's very slow to retrieve objects with many links to repository objects.
- Will investigate this week and create JIRA tickets when I get to specific issues.