...
- Nick Ruest
- Andrew Woods
- Yinlin Chen
- Daniel Lamb
- Colin Gross
- Danny Bernstein
- Bill BrananDavid Wilcox
- Youn Noh
- Osman Din
- David Wilcox
- Noah Botimer
...
Agenda
Update on graphs and summaries of completed tests
If R-scripts are ready, would be good for all to run over their results
Update on release testing participation
- UCSD? UT Austin? Princeton?
- Establishing environment baselines: Colin Gross
- How can the team help?
- Short-running tests?
- ...
Minutes
Update on graphs and summaries of completed tests
...
...
...
...
- grosscol/fcrepo_perf_analysis
- Script designed to work against the log output of the JMeter tests
- One JMeter component (that outputs a graph) should not be configured while running a test due to the performance impact (and the graph output is not needed)
- It would be useful if everyone were familiar with the R scripts, bringing R environments up to date, etc.
- Action: Everyone should run the scripts in local R environment to build familiarity
- Final test outputs can be attached to the test result pages
- We will need to distill these results into an overall message
Update on release testing participation
- A number of institutions are committed to testing future Fedora releases on an ongoing basis using their own production data
- This can have an impact on performance and scale testing as some institutions will have large collections/datasets
- We should add columns for committed institutions to the release testing pages so they can sign off on each release after testing
- We will continue to reach out to more institutions to add to the list
Establishing environment baselines
- Colin put together a script that identified characteristics of a test platform
- Andrew put together a script to identify additional performance characteristics
- Colin will refine JMeter tests to be short running rather than running until Fedora fails
- We can run these while varying characteristics along the way (I/O, memory, etc.)
- Establish characteristics between hardware and application performance
...
Actions
- Colin Gross will look into putting together a script for baselining hardware and network characteristics to be factored into each test run.
- Danny Bernstein: Summarize the existing test data.