[I sent the reply below a month ago, but I apparently only sent it to Marco, no the list.] 04/10/2012 08:15 PM, Marco Amadori: > The snapshot was required to not stress to much flash memory I'm doubtful of this claim. Snapshots are synced back in full during *each* shutdown (not even counting manual syncs) even if the persistent files they handle wasn't touched at all (and no new files were added). Hence, a snapshot is flash-friendlier than a bind if and only if the snapshot size is less than the amount of persistent data written during a single session. Compression for the snapshots helps in its favour, but we're still only talking about a factor of (say) 2 most of the time. To me it seems highly work-flow dependent on whether your claim is true or not, and I would almost be so bold to say that it's false for most common work-flows, but I'll refrain from doing it since I have no statistics to back it up with :). OTOH, since snapshots doesn't scale well (since their extracted contents are stored in RAM) they are only usable for smaller amounts of data, so one could make the case that my above "analysis" is too abstract and general for it to capture the reality of how snapshots are used. But if it's the case that snapshots should only be used for small amounts of data the increased flash wearing of using binds (even when completely rewriting all of the persistent data each session) would also be negligible. I believe "small amounts of data" (in the previous paragraph) must mean stuff that's not expected to grow that much (e.g. configuration files, keyrings). Stuff that doesn't grow much generally doesn't change (in place, so to speak) either. Hence binds seem like better choice. I guess where I'm headed with all this is that snapshots seems obsolete to me. In most cases binds are the better choice. Maybe we don't need snapshots any more? Or have I missed some realistic use case where binds actually wear out flash-based storage quicker than snapshots? Just some food for thought. Cheers!
Attachment:
signature.asc
Description: OpenPGP digital signature