SAP-C02 Question 316
Select 2Your organization runs a high-performance web application that requires low latency and high throughput for its in-memory caching layer. At the same time, you also need to process complex Machine Learning (ML) workloads that involve large matrix computations. Which instance families should you choose to optimize cost and performance for these two distinct workloads?
- A
R5 for the in-memory caching layer and P4 for Machine Learning workloads
- B
C6g for the in-memory caching layer and G5 for Machine Learning workloads
- C
M5 for the in-memory caching layer and C6i for Machine Learning workloads
- D
X2idn for the in-memory caching layer and P3 for Machine Learning workloads
- E
R6g for the in-memory caching layer and Inf1 for Machine Learning workloads
Show answer and explanation
Correct answers: A, E
Explanation
Choosing the right EC2 instance family is critical for optimizing cost and performance. R5 and R6g instances are memory-optimized, making them excellent choices for in-memory caching. P4 instances offer state-of-the-art GPUs for demanding ML training and inference tasks, while Inf1 instances provide a cost-effective solution for ML inference using AWS Inferentia chips. Selecting the appropriate instance family ensures that resources align with workload requirements without incurring unnecessary costs.
- A. Correct.
Correct. R5 instances are optimized for memory-intensive workloads like in-memory caching, and P4 instances are GPU-based, designed for high-performance ML training and inference.
- B. Incorrect.
Incorrect. While C6g instances are cost-effective for compute-intensive workloads, they are not ideal for in-memory caching. G5 instances are suitable for GPU-based workloads but may not be as efficient as P4 for ML training.
- C. Incorrect.
Incorrect. M5 is a general-purpose instance family that is not optimized for memory-intensive workloads like caching. Additionally, C6i is a compute-optimized family, not designed for ML workloads.
- D. Incorrect.
Incorrect. X2idn instances are optimized for high memory and storage but are overkill for in-memory caching. P3 instances are an older GPU-based instance family; P4 is a better choice for modern ML workloads.
- E. Correct.
Correct. R6g instances are memory-optimized and cost-effective for in-memory caching. Inf1 instances are designed specifically for ML inference and are more cost-efficient than GPU-based instances for such workloads.