Load Balancing and Auto Scaling Questions and Answers

21.

Match each scaling approach with when it suits.
List I
A. Reactive on a metric
B. Scheduled
C. Predictive
D. Manual
List II
1. repeating patterns learnt from history
2. gradual and unpredictable load changes
3. a planned event with a known large step
4. known patterns such as business hours

A.A-2, B-4, C-1, D-3
B.A-4, B-2, C-1, D-3
C.A-2, B-4, C-3, D-1
D.A-3, B-1, C-4, D-2

Answer: A

Most real systems combine several, with a schedule for the known pattern and a reactive policy for what the schedule did not anticipate. Relying on one alone is usually a mistake. The schedule handles the known load and the reactive policy handles the surprise.

Discuss in Forum

22.

Scaling in should be more conservative than scaling out because

A.the load balancer prevents any instance being removed
B.removing capacity too eagerly risks an outage if load returns
C.instances cost nothing at all while they sit idle
D.scaling in is technically much harder than scaling out

Answer: B

The cost of being slightly overprovisioned is small and predictable, while the cost of being short is an outage. Asymmetric thresholds are the usual arrangement. Being a little overprovisioned is a cost; being underprovisioned is an incident.

Discuss in Forum

23.

Assertion (A): Scale out and scale in thresholds are usually set asymmetrically.
Reason (R): The consequence of scaling in too far is far worse than that of scaling out too readily.

A.A is true but R is false
B.A is false but R is true
C.Both A and R are true, and R is the correct explanation of A
D.Both A and R are true, but R is not the correct explanation of A

Answer: C

Both are true and the reason gives the asymmetry of the risks, and it also happens to reduce flapping by widening the band between the two actions. One adjustment therefore addresses both the risk and the oscillation together.

Discuss in Forum

24.

Three of the following make an application easier to scale horizontally. Which is the odd one out?

A.holding session state in a shared external store
B.making request handling entirely stateless
C.externalising configuration out of the image
D.storing uploaded files on the instance disk

Answer: D

Files on a local disk are visible to one instance only and lost when it is replaced. The other three all remove state from the instance where it would prevent scaling. Object storage is the usual destination for the uploads in the fourth option.

Discuss in Forum

25.

State whether the following is true or false.
There is a point beyond which adding instances no longer improves throughput.

A.True
B.False
C.True only at large scales
D.False only at small scales

Answer: A

A shared database, a lock or a network path eventually becomes the constraint. Finding what saturates first is the whole business of scalability testing. The answer is rarely the component anyone expected before the test was run.

Discuss in Forum

26.

Assertion (A): Adding instances stops helping once a shared resource becomes the bottleneck.
Reason (R): Every instance contends for the same database, lock or network path.

A.A is false but R is true
B.Both A and R are true, and R is the correct explanation of A
C.Both A and R are true, but R is not the correct explanation of A
D.A is true but R is false

Answer: B

Both are true and the reason gives the mechanism, and the remedy lies in the shared resource rather than in adding yet more instances, which only increase the contention for it. Sharding, caching or a read replica addresses the cause where more instances cannot.

Discuss in Forum

27.

Arrange the steps of diagnosing a scaling problem in order.
1. address the shared bottleneck or add instances accordingly
2. establish which resource saturates first under load
3. verify with a load test that the limit has moved
4. determine whether it is per instance or shared

A.4 - 2 - 1 - 3
B.2 - 1 - 4 - 3
C.2 - 4 - 1 - 3
D.3 - 1 - 4 - 2

Answer: C

Adding instances without knowing what saturates is guesswork and often makes a shared bottleneck worse. The verification step is what confirms the diagnosis was right. Skipping it leaves you unable to say whether the change helped or merely coincided.

Discuss in Forum

28.

Consider the following statements about load testing.
1. It establishes where the system saturates before users find out.
2. It should exercise the same paths that real traffic uses.
3. A test against a scaled down environment predicts production exactly.

A.3 only
B.1 only
C.All of them
D.1 and 2 only

Answer: D

The first two are correct. The third is not so, since shared bottlenecks often appear only at scale and a small environment may show none of them. A test environment a tenth the size can be actively misleading about the limit.

Discuss in Forum

Loading…

Topics

Other Subjects

Tags: Load Balancing and Auto Scaling Questions and Answers || Load Balancing and Auto Scaling MCQ Questions and Answers || Load Balancing and Auto Scaling GK Questions and Answers || Load Balancing and Auto Scaling GK MCQ Questions || Load Balancing and Auto Scaling Multiple Choice Questions and Answers || Load Balancing and Auto Scaling GK || GK on Load Balancing and Auto Scaling || Cloud Computing Questions and Answers || Cloud Computing MCQ Questions and Answers || Cloud Computing GK Questions and Answers || GK on Cloud Computing

Load Balancing and Auto Scaling Questions and Answers

Topics

Other Subjects

About Us

Stay informed about all important Dates & Notifications