There's lot's good explanations in the documentation on how to backup your indexes.
I haven't looked at Splunk5 yet, but unless they've changed things significantly from Splunk 4, it's mostly based on size and number of files, not date, so you need to look at what you are collecting, and then tune your limits for the various bucket types to match your requested retention period. You can also manually force a roll from hot to warm, if you want to guarantee that buckets are backed up at given times.
And yes: The exception is the frozenTimePeriod, which I understand does what you're asking: it will roll data to frozen (delete or archive) after a given number of seconds.