1Z0-1072-25 Question 4
Single answerYou are designing an HPC (High Performance Computing) application on Oracle Cloud Infrastructure (OCI) that performs iterative computations on a large, ever-changing dataset. The intermediate results do not need to be retained once processing is complete, and you require extremely low-latency local storage to optimize performance. Which compute configuration should you choose to meet these requirements?
- A
Use a DenseIO instance shape with local NVMe storage for ephemeral data
- B
Use a standard instance shape with attached OCI block volumes
- C
Use a GPU instance shape and store all data on OCI Object Storage
- D
Use a flexible instance shape with increased memory allocation for caching
Show answer and explanation
Correct answer: A
Explanation
When designing HPC solutions on OCI that require extremely fast, temporary storage, DenseIO shapes with local NVMe drives are typically the best choice. They offer high IOPS and very low latency, crucial for iterative computations on ephemeral data. For more details, consult the Oracle Cloud Infrastructure documentation on Compute shapes and storage options: https://docs.oracle.com/en-us/iaas/Content/Compute/References/computeshapes.htm
- A. Correct.
Correct. DenseIO shapes include locally attached NVMe storage, which provides high IOPS and low latency for ephemeral workloads. Since the dataset is not needed after processing, using local NVMe drives is ideal for high-performance temporary storage needs.
- B. Incorrect.
Incorrect. Standard shapes rely on block volumes for additional storage, which are persistent network-attached volumes. While OCI block volumes are reliable, they do not match the ultra-low latency of locally attached NVMe drives, which is crucial for HPC workloads.
- C. Incorrect.
Incorrect. GPU shapes provide specialized hardware accelerators but do not necessarily include high-performance local NVMe storage by default. Storing ephemeral data in Object Storage will introduce higher latency compared to locally attached NVMe.
- D. Incorrect.
Incorrect. Flexible shapes allow you to customize the OCPU and memory count, but this option alone does not provide ultra-low-latency, locally attached storage. Increasing memory can help with caching, but it does not replace the high IOPS required for HPC ephemeral datasets.