Q: 7
You are a program manager within a Software as a Service (SaaS) company that offers rendering
software for animation studios. Your team needs the ability to allow scenes to be scheduled at will
and to be interrupted at any time to restart later. Any individual scene rendering takes less than 12
hours to complete, and there is no service-level agreement (SLA) for the completion time for all
scenes. Results will be stored in a global Cloud Storage bucket. The compute resources are not bound
to any single geographical location. This software needs to run on Google Cloud in a cost-optimized
way.
What should you do?
Options
Discussion
I think A makes more sense here, since preemptible instances are much cheaper and work well when your jobs can handle interruptions. The workload isn't tied to SLAs and each scene is under 12 hours, so preemptibles fit perfectly. Not totally sure though if D would ever beat A unless there's a strict job concurrency need. Agree?
D imo. Starting more instances with fewer vCPUs should let you parallelize the scenes and keep jobs flexible for interruption, which seems to fit. A similar scenario popped up in some practice tests, but official docs recommend preemptibles more often. Anyone else stick with D after reviewing guides?
Be respectful. No spam.
