Horizontal scaling means
Answer: A
Vertical scaling means moving to a bigger machine and eventually runs out of machine. Horizontal scaling has no such ceiling but requires the application to be built for it. Most cloud designs therefore favour many modest instances over one large one.
Please log in or register to report this question.
Assertion (A): Horizontal scaling requires the application to be designed for it.
Reason (R): State held in one instance is not visible to the others.
Answer: B
Both are true and the reason gives the requirement, and moving session state to a shared store is the usual first step in making an application scale out. Sticky sessions are a workaround for having skipped that step rather than a design.
Please log in or register to report this question.
Consider the following statements comparing horizontal and vertical scaling.
1. Vertical scaling is limited by the largest available instance.
2. Horizontal scaling requires the workload to be distributable.
3. Horizontal scaling works for any application without change.
Answer: C
The first two are correct. The third is precisely the difficulty, since a stateful monolith cannot simply be run in several copies behind a balancer. The work of making it scale out is largely the work of removing local state.
Please log in or register to report this question.
Match each load balancing algorithm with what it does.
List I
A. Round robin
B. Least connections
C. Weighted
D. IP hash
List II
1. backends receive traffic in set proportions
2. each backend in turn regardless of load
3. the same client always reaches the same backend
4. the backend with fewest open connections
Answer: D
The choice matters when backends differ in capacity or when requests differ greatly in cost. Round robin is the default and is adequate for uniform backends and uniform requests. Once either assumption fails, a load aware algorithm distributes work far better.
Please log in or register to report this question.
Least connections balancing suits workloads where
Answer: A
Round robin would keep sending work to a backend already tied up with a long request. Counting open connections approximates the load each backend is actually carrying. It is a cheap proxy that needs no cooperation from the application itself.
Please log in or register to report this question.
Assertion (A): Round robin can distribute load unevenly even though it distributes requests evenly.
Reason (R): Requests differ in cost, so an equal count does not mean an equal load.
Answer: B
Both are true and the reason gives the distinction, and it is why least connections or a load based algorithm is preferred for heterogeneous work. A single slow endpoint can otherwise monopolise one backend while others idle.
Please log in or register to report this question.
A load balancer health check exists to
Answer: C
A failing backend is removed from the pool until it recovers. The check must exercise something meaningful, since one that always succeeds provides no protection at all. A check that queries the database it depends on catches far more real failures.
Please log in or register to report this question.
Consider the following statements about health checks.
1. A check should exercise the dependencies the request path needs.
2. Too aggressive a check can remove healthy backends under load.
3. A check that always returns success is adequate.
Answer: D
The first two are correct and describe the balance to be struck. The third is exactly the failure mode, since such a check will happily route traffic to a broken backend. Depth of check and cost of check pull against each other and must be balanced.
Please log in or register to report this question.
Assertion (A): A health check that is too aggressive can make an overload worse.
Reason (R): It removes backends that are merely slow, concentrating the load on those that remain.
Answer: A
Both are true and the reason gives the failure mode, and the resulting cascade can take out a whole pool that would otherwise have degraded gracefully. Slow but serving is nearly always better than removed from the pool entirely.
Please log in or register to report this question.
Session affinity in a load balancer sends
Answer: B
It allows a backend to hold session state in memory but undermines even load distribution. Moving the state to a shared store removes the need for it entirely. Affinity should be treated as a temporary accommodation, not a permanent design.
Please log in or register to report this question.
