Meetings are pending with mu2e and g-2. Preferably schedule for early next week.
Robert has a fresher copy of his requirements document. He emailed it to everyone.
We should put the requirements in a wiki so we can all hack on it.
A campaign consists of more than one task. Tasks are the individual submissions.
A campaign example:
Process all raw data with version x reco and version y summary. Because this may be huge we need to generate subsets of the dataset.
We don't want to keep these logs in this database.
Perhaps have web services for this. A call can be something like "get me logs for this project name".
If we get the logs from job sub for a 3 day range we should be alright.
Long term archiving is a separate project. (Robert)
If you use SAM, then you get added functionality in the system. But not all projects need to rely on SAM, so this is OK, you just won't get the added functionality.
SAM dataset names
SAM queries (sam dimensions)
This will be a multi tenant application; scale vertically.
It is common to have to go back 9 months ago and look at logs. We can archive logs in in a separate readonly partition.