Cloud Storage Systems Questions and Answers

Answer: C

Lifecycle rules act on where and whether an object is stored, never on what it contains. The other three are exactly the actions such rules perform. Placement rather than content is the concern. Contents are never altered by a rule.

Answer: D

Eventual consistency permits a window in which a read may see older data. Modern object stores mostly offer strong consistency, which removed a long standing source of application bugs. Eventual consistency allowed a stale window.

Answer: A

Both are true and the reason gives the difficulty, and the workarounds people wrote for it are a large part of why strong consistency was so welcomed. Read after write is the case that bites. The window was short but real.

Answer: B

Correct answer: A-2, B-4, C-1, D-3

Four mechanisms answering different risks. Synchronous replication acknowledges only when every copy is written, so nothing is lost but the write waits for the slowest replica. Asynchronous acknowledges first and copies after, which is faster but leaves a window in which recent writes can be lost. Erasure coding reconstructs data from fragments and parity, giving durability at far less overhead than whole copies. And cross region replication protects against losing an entire region.

Answer: C

A scheme storing ten data and four parity fragments survives four losses for forty per cent overhead rather than the two hundred of three copies. The cost is more computation and slower reconstruction of a lost fragment.

Answer: D

Correct answer: 40 per cent

\[ \frac{4\ \mathrm{parity}}{10\ \mathrm{data}} = 0.40 = 40\%, \] so 14 fragments are stored for 10 of data. That is the attraction of erasure coding: triple replication costs 200 per cent overhead for the loss of two copies, while this scheme tolerates the loss of any four fragments for 40 per cent. The cost is computation on every read that must reconstruct, and higher latency when fragments are missing.

Answer: A

Both are true and the reason gives the economics, and the extra computation involved is a good trade when the alternative is buying three times the disks. Durability is comparable in both. Computation replaces raw capacity.

Answer: B

Correct answer: protects the media but not a valid credential

Encryption at rest defends against the physical layer - a stolen or decommissioned disk, a mislaid backup. It does nothing against a request bearing valid credentials, because the service decrypts on the way out as a matter of course. Access control, key management and audit are what defend against that, which is why encryption at rest is a compliance baseline rather than a security strategy.

Answer: C

The first two are correct. The third is the most common misunderstanding, since the service decrypts transparently for anyone whose request is properly authorised. Access control is the other half. Media disposal is what it defends against.

Answer: D

Disabling the key renders the data unreadable even to the provider's services. It also means losing the key loses the data, which is a genuine operational risk. Revocation becomes possible with it. Key loss is then the customer's problem.