3 ms·
I've used them for a few years, not been particularly impressed with their speed. I'm also currently fighting a problem with the client quietly wedging on uplo
by Freaky 9y ago
I've used them for a few years, not been particularly impressed with their speed. I'm also currently fighting a problem with the client quietly wedging on upload (not sure how long ago it stopped, but the upload queue was past 6GB), and the re-syndication process just sitting there seemingly doing nothing.
> Uses aggressive de-duping on their backend (ZFS based I think?)
Err, I hope not. Pretty sure they sync a database of seen blocks between clients and they collectively dedupe based on that - their backend simply shouldn't have the information necessary to dedupe itself.
- corford 9y ago>Err, I hope not. Pretty sure they sync a database of seen blocks between clients and they collectively dedupe based on that - their backend simply shouldn't have the information necessary to dedupe itself. Think you're right based on reading: https://spideroak.com/faq/what-is-deduplication https://spideroak.com/faq/what-is-deduplication I assumed they'd use both techniques (i.e. de-dup client side when possible to save on upload) but that the way in which encrypted blocks were produced before uploading was deterministic (a bit like a hash function) so that if the same file was marked for backup on a different device the resulting blocks would be the same and could be de-duped server side. Providing a unique salt was used client side (shared across all devices and not known by Spideroak) while making the block, it should be possible to end up with deterministic encrypted blocks that are still unique outside of your account (I think anyway... barely know what I'm talking about when it comes to the intricacies of crypto!)