There are some parts of the CronArchive algorithm that seem inefficient. We could probably optimize and simplify the algorithm with the following changes:

• Instead of running archiving for lastN days/weeks/months/etc., the date range to initiate archiving should be determined (either by last archiving time or by command line option) and the optimal archiving requests should be made. It makes no sense to archive for last 50 days and then archive for the last 2 years.
  • Based on the date range, day archiving for every day in the range should be scheduled individually, then week, then month, then year. This way, period archiving will not actually initiate archiving for sub-periods and every individual archiving request can be parallelized.
• Implement bottom up aggregation for period archiving (ie, mimic incremental map reduce). Right now, when cron archive is run again, the year, month, week and day periods for today are re-archived completely. That is to say, when archiving for a year, the data for every month is accessed and aggregated in PHP. This is slow. Instead, we can take the new data for the current day and add it to existing week + month + year periods. May run into issues w/ report limiting features.