1Z0-1072-25 Question 10
Single answerYour organization needs to run large-scale computational fluid dynamics (CFD) simulations that require fast inter-node communication, high-performance computing libraries, and low-latency networking. In addition, you want ephemeral local storage for scratch data. Which Oracle Cloud Infrastructure (OCI) compute shape would be the most appropriate choice for this workload?
- A
VM.Standard3.Flex
- B
BM.GPU3.8
- C
BM.Standard3.16
- D
BM.HPC2.36
Show answer and explanation
Correct answer: D
Explanation
HPC shapes such as BM.HPC2.36 offer specialized hardware configurations, including RDMA-based cluster networking for low latency between nodes, which is essential for HPC workloads like CFD simulations. They also include local NVMe storage for high-speed temporary data access. For more details, refer to the Oracle Cloud Infrastructure documentation on HPC shapes and best practices for running distributed simulations.
- A. Incorrect.
Option 1 (VM.Standard3.Flex): While flexible shapes allow you to configure OCPUs and memory, they are not optimized for tightly coupled HPC workloads with low-latency interconnect. They also do not provide dedicated HPC networking features.
- B. Incorrect.
Option 2 (BM.GPU3.8): This shape provides GPU acceleration, which is suitable for workloads like machine learning or high-end graphics. However, traditional CFD simulations often benefit more from HPC-specific CPU-based resources with RDMA networking, making a GPU-focused shape not the best fit.
- C. Incorrect.
Option 3 (BM.Standard3.16): This bare metal shape offers dedicated resources but lacks the specialized RDMA-based cluster networking features that HPC shapes provide for scaling computational workloads across multiple nodes.
- D. Correct.
Option 4 (BM.HPC2.36): This is the correct choice. HPC shapes in OCI are designed for highly parallel, compute-intensive tasks. They include low-latency RDMA networking and local NVMe storage, making them optimal for CFD simulations and other HPC cluster workloads.