Previous Action Items

Status Reports

Analyzing new data set uploaded yesterday
- the test for equality of variance showed equal variance for good and broken build (for this particular break)
- Autocorrelation check for 1 reboot 100 data points: dropped off the first data point, and looked at the rest. Discovered that every data point (except the first) is uncorrelated. Which means, we can take any data point without worrying that we are duplicating information.

Config changes: new section for pageloader options in config files so we can override things like 'tpmanifest' and 'tpcycles' from PerfConfigurator or run_tests

Lots of config changes to talos
Looking to merge pageloader into talos codebase
Have a good setup for running requests from metrics
- Will start merging these changes into pageloader/talos as necessary
Config changes: ability to override manifest, cycles, other pageloader options from command line. Additional clean up to make configuration more flexible for future changes.

Great Graphserver Unifcation - "mr. helmer, tear down this wall"
- bug 721895 - merge hg/git codebases
- bug 721884 - move writes from graphs-old to graphs.m.o
- currently in-testing, should be almost done here

rhelmer arranged for IT to create a copy of the graphs_mozilla production database on the stage machine. Jeads copied this locally and setup a local development environment.
Jeads worked on adapting the bughunter javascript and webservice to work in this environment to implement the data view concept proposed in the mocks.
The production database is 1.7GB. Backfilling a new schema with the existing data will not be a problem with this quantity of data.
Not sure what the growth rate of the data is but unless it's predicted to grow exponentially MySQL will be more than sufficient to manage the new version of the schema.

What should http://graphs.mozilla.org/graph.html look like?
- branches should be more sane
- should be able to select by test (really, manifest), chrome/nochrome, event (onload, mozafterpaint), etc

bug 706912 - implement regression calculations - A*Team
collect a smaller regression that would not be caught or so obvious with 100 cycles then reboot
disable side by side staging - armenzg, jmaher