D is correct

ans: D

It seems A & D both as suitable answer. But I'll go with A, as attaching GPU with preemptible VM will increase cost, and in this question the purpose of opting for preemptible VM is reducing cost. https://cloud.google.com/compute/docs/instances/preemptible#preemptible-with-gpu

Will go with D. Looking for cost effective. https://cloud.google.com/compute/docs/instances/preemptible#preemptible-with-gpu

ans: A Preemptible VMs are best suited for fault-tolerant, non-critical applications due to their temporary nature. Among the options listed, A, a scalable in-memory caching system, aligns well with preemptible instances as it can handle interruptions and doesn't require continuous uptime.

Option D

I would go with A compared to other options

Ans is D Video rendering service is like application type called Batch job. Therefore, we can use instance type preemptible for them. If they complete task, they could be destroy and generate new instance to work continuously next task.

C is more accurate

Why not A? "A scalable in-memory caching system." In general a caching system is not critical to the function of an application. If the cache is down it will cause requests to have cache miss and query the DB instead. User requests will still get served albeit slower. In addition the answer specifies that the caching system is "scalable" reducing further the impact of 1 VM getting preempted, ie traffic can be automatically redirected to other cache replicas. To me all other answers seem to have a more severe impact on the user in case the VM is preempted.

C is the right answer

i think is C, database cluster is storage, and distributed, eventually consistence is resistant for the preempted. and sufficient quorum can ensure the DB transitions.

D is correct

D is correct

D https://cloud.google.com/compute/docs/instances/preemptible

D - external persistent storage