Free AWS AIP-C01 Actual Exam Questions - Question 14 Discussion
Scenario: Autonomous vehicle model training experiences slow startup times and low GPU utilization because the training job downloads data sequentially from S3. Goal: improve data access performance and training throughput while maintaining the S3 repository and avoiding data duplication. Question- Which solution should be implemented to optimize SageMaker AI training performance while maintaining the existing S3-based workflow?. Options:
It’s D, but it could cause delays since copying data every run wastes time. A is better because FSx for Lustre caches data and avoids repeated downloads, which directly improves GPU utilization.
A/C? I’m going with A here, because FSx for Lustre is designed to handle high-throughput workloads and can cache data locally, which should boost GPU usage. C might help a bit by speeding up data transfer over the network, but it won’t fix the sequential download bottleneck as effectively. B’s EFS isn’t optimized for big, streaming data like this, and D sounds like it’d just add overhead with copying each time. So A seems like the best balance of performance improvement and maintaining the S3 workflow.
A/C? A seems best for throughput, but enabling S3 Transfer Acceleration (C) could also reduce latency when fetching data, speeding up startup without changing storage setups much.
Yeah, option A seems solid because FSx for Lustre handles high throughput from S3 without duplicating data, unlike D which involves slow copying. B and C don't really solve the sequential download bottleneck. A
A imo, since FSx for Lustre lets you stream data fast without copying or changing the S3 setup.
It’s A, because FSx for Lustre caches S3 data and provides high-throughput access, improving GPU utilization without duplicating data or copying it locally like in D. It’s a better fit than EFS or Transfer Acceleration here.
D imo, bar charts can’t show size well, so scatter with color and size fits best.
Probably A, since Data Wrangler scatter plots also handle color for dimensions well.
Maybe D is right because scatter plots are perfect for showing two numeric axes plus color and size for extra info. The other options don’t seem to handle all four dimensions as clearly.
D, since scatter plots naturally handle both color and size mappings well.
D seems like the best fit since it uses scatter plot with color and size to show all four dimensions clearly.