• No I still don't understand; unless I've completely misunderstood, in the scenario you provide the value on the row changes, meaning that value now should be in a different partition from when it was inserted, i.e. it's now two days old rather than one. The data is effectively updated at that point (the partition value 1 needs to be rewritten as value 2, the page has to come off disk into buffer be changed, and put back to disk), which would mean it has to be rewritten to the new partition?

    I (...think I...) understand the metadata change for partition switching, but in the scenario above the actual data row itself changes to move it from the first to the second partition? If the data was static and new partition ranges were added, then the underlying data would never have to move between partitions.

    More so, in my scenario where I have a wider window, only some of the data would move between the partitions; (e.g. with chunks of data of three months, at month change, only the oldest month would be moved), so there could even be data at a page level that would need to exist within different partitions?

    Follow me on twitter @EvoDBACheck out my blog Natural Selection DBA[/url]