Oracle Cloud Infrastructure | Optimize Performance for AI/ML Applications

How to Get GPU and 20Gbps Network Throughput for Your AI/ML Application

Question

You have an AI/ML application running on Oracle Cloud Infrastructure.

You identified that the application needs GPU and at least 20Gbps Network throughput.

The application is currently using a VM.Standard2.1 compute without any block storage attached to it.

Which two options allow you to get your required performance for your application? (Choose two.)

Answers

Explanations

Click on the arrows to vote for the correct answer

A. B. C. D. E.

DE.

Sure, I'll explain each option and its suitability for the requirements mentioned in the question.

Option A: Terminate the compute instance preserving the boot volume. Create a new compute instance using the BM.GPU2.2 shape using the boot volume preserved, but no block volume attached.

Explanation: The BM.GPU2.2 shape is a bare metal instance that comes with two NVIDIA Tesla V100 GPUs, which are ideal for AI/ML workloads that require high-performance computing power. This option also preserves the boot volume from the existing compute instance, which saves time and effort in reconfiguring the new instance. However, the option does not provide the required network throughput of 20Gbps.

Conclusion: Option A provides the required GPU resources but not the required network throughput. Therefore, this option is not suitable for the given requirements.

Option B: Terminate the compute instance preserving the boot volume. Create a new compute instance using the VM.Standard2.2 shape using the boot volume preserved, but no block volume attached.

Explanation: The VM.Standard2.2 shape is a virtual machine instance that comes with 1 GPU and 36 vCPUs. This option preserves the boot volume from the existing compute instance, which saves time and effort in reconfiguring the new instance. However, this option does not provide the required network throughput of 20Gbps.

Conclusion: Option B provides a GPU resource but not the required network throughput. Therefore, this option is not suitable for the given requirements.

Option C: Terminate the compute instance preserving the boot volume. Create a new compute instance using the VM.GPU3.4 shape using the boot volume preserved and use the NVMe devices to host your application.

Explanation: The VM.GPU3.4 shape is a virtual machine instance that comes with 1 GPU and 36 vCPUs. It is designed for workloads that require high-performance computing power, such as AI/ML. This option also allows using NVMe devices to host your application, which can provide faster and more reliable storage for your application. This option satisfies both the required GPU resources and network throughput of 20Gbps.

Conclusion: Option C satisfies the given requirements of GPU resources and network throughput, and also provides a better storage option. Therefore, this option is suitable for the given requirements.

Option D: Terminate the compute instance preserving the boot volume. Create a new compute instance using the BM.HPC2.36 shape using the boot volume preserved and use the NVMe devices to host your application.

Explanation: The BM.HPC2.36 shape is a bare metal instance that comes with 2 Intel Xeon Platinum 8167M processors and 2 Mellanox ConnectX-5 100 Gb/s NICs. It is designed for workloads that require high-performance computing power and network throughput, such as HPC and AI/ML. This option also allows using NVMe devices to host your application, which can provide faster and more reliable storage for your application. This option satisfies both the required GPU resources and network throughput of 20Gbps.

Conclusion: Option D satisfies the given requirements of GPU resources and network throughput, and also provides a better storage option. Therefore, this option is suitable for the given requirements.

Option E: Terminate the compute instance preserving the boot volume. Create a new compute instance using the BM.GPU2.2 shape using the boot volume preserved and attach a new block volume to host your application.

Explanation: The BM.GPU2.2 shape is a bare metal instance that comes with two NVIDIA Tesla V100 GPUs, which are ideal for AI/ML workloads that require high-performance computing power. This option also allows attaching a new block volume to host your application. However, this option does not provide the required network throughput of 20