Professional Cloud ArchitectDemo

By Google · Question Mode

QuestionQ284

Designing and planning a cloud solution architecture

Your company runs AI/ML workloads on graphics processing units (GPUs) in Google Kubernetes Engine (GKE). These workloads are deployed across production and development projects, and the production environment is business-critical and runs 24/7. You must ensure GPU capacity is always available during scaling events without being constrained by overall GPU availability. You also need to be able to shift unused GPU capacity from the production project to development projects as necessary. You want to follow Google-recommended practices. What should you do?

  • A Create a reservation in the production project, and then transfer the reservation to the development project when needed.
  • B Create a shared reservation in the production project, and then add the development project as a consumer project to move capacity when needed.
  • C Create an owner project with shared reservations, and then configure the production project and development project as consumer projects to consume capacity.
  • D Create a reservation in the production project, and then create a separate reservation in the development project to secure capacity.
Explanation

Shared reservations guarantee capacity for matching GPU VM resources and can be consumed by multiple designated projects. Google recommends an owner project that creates and manages the shared reservations but does not consume them, with workload projects configured as consumer projects. This lets production and development draw from the same reserved capacity, maximizing use of capacity that production is not currently consuming.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!