Cloud Storage Systems Questions and Answers

Answer: A

Correct answer: addressed by key and replaced whole

An object store is a flat namespace of keys mapping to blobs, and an object is written or replaced in its entirety rather than edited in place. Block storage presents raw blocks that a filesystem addresses by offset and modifies individually. That single difference explains everything else: object stores scale almost without limit and are reached over HTTP, but they cannot serve as the disk under a database.

Answer: B

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

Four storage types for four access patterns. Object storage suits large, essentially immutable items - backups, media, logs. Block storage is the disk for a filesystem or database on one instance. File storage provides a shared filesystem that several instances mount at once. And an instance store is fast local disk that vanishes when the instance stops, so it is for scratch data only.

Answer: C

Both are true and the reason gives the requirement, and object storage suits the database's backups perfectly well even though it cannot hold the live files. Whole object replacement is the rule. Backups are a natural fit for it.

Answer: D

The first two are correct. The third is exactly what it does not do, and an application that expects to seek and write within a file cannot use it directly. Replacement is the only update. Seeking within an object is impossible.

Answer: A

Correct answer: the probability that stored data is not lost

Durability and availability are different guarantees. Durability is about permanence - whether the bytes still exist - and is quoted as a number of nines per year. Availability is about reachability at a given moment. A service can be highly durable and briefly unavailable, which is the normal case during an outage: the data is intact and cannot be read for an hour.

Answer: B

Both are true and the reason gives the distinction, and a service may lose availability for an hour without any risk at all to the durability of what it holds. Both are quoted as nines. Neither figure implies the other at all.

Answer: C

Correct answer: far less than one

Eleven nines means an annual loss probability of \( 10^{-11} \) per object, so \[ 10^{7} \times 10^{-11} = 10^{-4}, \] one ten-thousandth of an object expected in a year. Put another way, ten million objects would be expected to survive about ten thousand years before losing one. At that level the realistic risks are deletion by mistake and account compromise, not media failure - which is why versioning matters more than durability figures.

Answer: D

The first two are correct and describe the trade. The third ignores retrieval cost, which can easily exceed the storage saving for data that is actually read. Access frequency decides the class. Storage price is only half the sum.

Answer: A

Both are true and the reason gives both traps, and moving data to a cold class that is then read regularly is a common way to raise a bill rather than lower it. Minimum durations apply as well. Early deletion is charged for in full.

Answer: B

Correct answer: moves or deletes objects as they age

A lifecycle policy acts on age: transition to a colder, cheaper class after so many days, expire after so many more, and clean up old versions and incomplete uploads. It is the standard mechanism for controlling storage cost without anyone having to remember, and it is quite separate from access control, encryption and replication.