1Z0-1072-25 Question 9
Single answerYour company needs to run large-scale scientific simulations that demand high-performance computing (HPC) capabilities, low-latency RDMA cluster networking, and ephemeral local NVMe storage to handle intensive parallel file I/O. Which Oracle Cloud Infrastructure compute shape is the most appropriate choice for this scenario?
- A
VM.Standard.E4.flex
- B
BM.Standard.E4.128
- C
BM.GPU3.8
- D
BM.HPC3.36
Show answer and explanation
Correct answer: D
Explanation
For high-performance computing workloads�especially those involving numerous simultaneous simulations, large data transfers, and intensive parallel processing�an HPC-optimized shape with RDMA networking and local NVMe storage is critical. BM.HPC3.36 meets these requirements by providing the requisite CPU cores, HPC-optimized network interconnect, and fast ephemeral storage directly attached to the host. Refer to Oracle Cloud Infrastructure documentation on 'High Performance Computing Solutions' for more details on HPC shape characteristics and best practices.
- A. Incorrect.
Option 1 (VM.Standard.E4.flex): While manually adjustable CPU and memory resources can help optimize cost and performance for various workloads, VM.Standard.E4.flex does not provide RDMA-based cluster networking or built-in ephemeral local NVMe storage. It's not ideal for intensive HPC simulations that need low-latency interconnects and high-throughput local disk operations.
- B. Incorrect.
Option 2 (BM.Standard.E4.128): This bare metal shape offers a large number of CPU cores but does not feature HPC-specific RDMA cluster networking, nor does it come with ephemeral local NVMe storage. It would require attaching block volumes separately, which can add latency compared to local NVMe.
- C. Incorrect.
Option 3 (BM.GPU3.8): While GPU shapes are excellent for machine learning and compute-accelerated workloads, they do not specifically target the low-latency, HPC cluster networking offered by HPC shapes. They also may not include the same scale of ephemeral local NVMe necessary for large-scale HPC I/O needs.
- D. Correct.
Option 4 (BM.HPC3.36): This is an HPC-optimized bare metal shape that includes high-bandwidth RDMA cluster networking and ephemeral local NVMe storage, making it suitable for running large-scale simulations that require low-latency, high-throughput file operations. It is specifically designed for HPC workloads.