1Z0-1067-25 Question 45
Select 2You are managing a high-performance computing (HPC) environment on Oracle Cloud Infrastructure (OCI) that runs scheduled machine learning training jobs for four hours every day. The HPC cluster must meet strict performance requirements when jobs run, but it remains idle outside the scheduled window. Which two actions should you take to minimize costs without compromising performance?
- A
Enable instance scheduling to automatically start and stop HPC compute nodes outside of the scheduled job hours.
- B
Use a shape with minimal CPU and memory resources to continuously run the HPC cluster at a lower cost.
- C
Use OCI Flexible Shapes to right-size CPU and memory resources based on specific HPC workload demands.
- D
Migrate the HPC jobs to Always Free tier compute instances to eliminate ongoing compute costs.
Show answer and explanation
Correct answers: A, C
Explanation
To optimize cost and performance in HPC environments on OCI, you should schedule nodes to run only during the required window and leverage Flexible Shapes to precisely align resources to workload demands. This reduces unnecessary spend when nodes are idle and ensures sufficient CPU and memory allocation during training. For further guidance, refer to Oracle� cost optimization best practices and HPC design recommendations in the Oracle Cloud Infrastructure documentation.
- A. Correct.
Option 1: CORRECT. Enabling automated instance scheduling or using a function-based approach to start and stop nodes outside of job hours is an effective way to minimize idle runtime costs without affecting performance when training jobs are active.
- B. Incorrect.
Option 2: INCORRECT. Simply switching to the smallest shape would compromise performance for HPC workloads, since these jobs require sufficient CPU and memory resources to run efficiently.
- C. Correct.
Option 3: CORRECT. OCI Flexible Shapes allow you to adjust CPU and memory exactly to your HPC requirements, ensuring you pay only for the resources you need without sacrificing performance.
- D. Incorrect.
Option 4: INCORRECT. Always Free compute instances are limited in capacity and are not generally suitable for HPC workloads requiring high compute power, so this would likely fail to meet performance needs.