As the original answer mentioned, this is highly dependent on your storage type. Volume-level snapshots are a great way to save space on backups if you're comfortable with having all your data "online" without the option for offline backups. It's reasonably safe with an enterprise-level storage system, but be aware that you still have all your eggs in one basket. Our out-of-the-box backup scripts do a raw DB dump, which is the least space efficient method but it's simple, straightforward and guaranteed to restore nicely with a minimum of fuss. The problem is that as your data set grows, the time and effort required to shuffle around backups becomes problematic.
It's important to define your data retention period up front to get accurate space estimates. How long do you need the data? The typical answers are 30, 90 or 180 days, with 90 days being pretty common. For very large data sets, 30 days or less is typical. Note that when you prune detailed, per-transaction data from the analytics database you retain the aggregate reporting data, which may be enough to troubleshoot problems or analyze traffic patterns older than 30 days. Once you know the time horizon, you can plan for both storage needs and backup needs (if any).
Finally, since analytics data is only useful for a limited period of time, it may be an acceptable risk to build an HA analytics backend and simply rebuild if multiple analytics databases are completely lost. The risk of such an event happening is vanishingly small and accepting that risk can save greatly on storage costs for the very largest analytics databases.
LinkedIn
Twitter
