During a review, someone notes that object storage offers eleven nines of durability and concludes that no backup or retention strategy is needed for lakehouse tables. Is this reasoning sound?
Select an answer to reveal the explanation.
Short Explanation
Durability means the storage won't lose your bytes. It doesn't protect against someone dropping the wrong table or overwriting data. You need versioning, retention, and snapshot history for that.
Full Explanation
Durability guarantees describe the probability that stored objects survive hardware and infrastructure failure. They offer no protection against an erroneous or malicious request to delete or overwrite, which is the far more common cause of data loss. Object versioning, retention rules, and the snapshot history maintained by open table formats are the controls that address it.
The answer saying that durability of that level makes data loss statistically impossible is incorrect because it treats one failure mode as if it were all of them. The operation that removes the data is faithfully executed and perfectly durable.
The answer saying that durability applies only to objects under one gigabyte is incorrect as a factual claim; durability is not conditioned on object size.
The answer saying that no backup or retention strategy is needed provided the bucket is replicated to a second region is incorrect because cross-region replication propagates deletions as diligently as it propagates writes. It addresses regional availability, not mistakes.
For the exam, separate durability, availability and recoverability; questions in this area usually turn on the difference.