Question 230
An ML service must return predictions synchronously to application requests. The team is required to run a particular model, scoring code, and runtime configuration as a named implementation of one endpoint. Which option is the best fit?
- Endpoint request timeout
- Endpoint traffic allocation
- Managed online endpoint
- Deployment under endpoint
Explanation
Correct Answer
D:An online deployment packages inference assets behind an endpoint. Multiple deployments can support safe rollout without changing the client-facing endpoint name.
Incorrect Answers
A:Timeout settings bound how long requests can run before failing; extending them may mask slow scoring code and is not a substitute for profiling; the outstanding need is to run a particular model, scoring code, and runtime configuration as a named implementation of one endpoint.
B:Traffic weights support canary and blue/green testing among deployments; merely registering another model version does not change live request routing; this approach leaves the team unable to run a particular model, scoring code, and runtime configuration as a named implementation of one endpoint.
C:A managed online endpoint handles synchronous requests using deployed inference code and compute; it differs from batch processing that writes results for large offline inputs; the needed capability is to run a particular model, scoring code, and runtime configuration as a named implementation of one endpoint.
Question 231
An ML service must return predictions synchronously to application requests. An operator is tasked to route a controlled percentage of requests to a new deployment while preserving the old deployment. Which mechanism provides that capability?
- Online endpoint authentication mode
- Online deployment instance count
- Endpoint request timeout
- Deployment under endpoint
- Endpoint traffic allocation
Explanation
Correct Answer
E:Traffic weights support canary and blue/green testing among deployments. Merely registering another model version does not change live request routing.
Incorrect Answers
A:Endpoint authentication controls who may invoke the service; it does not replace identity permissions needed by scoring code to read external data; the requirement, by contrast, is to route a controlled percentage of requests to a new deployment while preserving the old deployment.
B:Replica count controls deployed serving capacity within available quotas; it does not alter model accuracy or evaluation-set construction; the team would still need to route a controlled percentage of requests to a new deployment while preserving the old deployment.
C:Timeout settings bound how long requests can run before failing; extending them may mask slow scoring code and is not a substitute for profiling; the specified outcome instead requires the team to route a controlled percentage of requests to a new deployment while preserving the old deployment.
D:An online deployment packages inference assets behind an endpoint; multiple deployments can support safe rollout without changing the client-facing endpoint name; the implementation under review needs to route a controlled percentage of requests to a new deployment while preserving the old deployment.
Question 232
An ML service must return predictions synchronously to application requests. The immediate design goal is to require callers to present a supported credential before receiving model predictions. Which approach addresses that goal?
- Online endpoint authentication mode
- Online deployment instance count
- Endpoint traffic allocation
- Managed online endpoint
Explanation
Correct Answer
A:Endpoint authentication controls who may invoke the service. It does not replace identity permissions needed by scoring code to read external data.
Incorrect Answers
B:Replica count controls deployed serving capacity within available quotas; it does not alter model accuracy or evaluation-set construction; the requested task remains to require callers to present a supported credential before receiving model predictions.
C:Traffic weights support canary and blue/green testing among deployments; merely registering another model version does not change live request routing; this alternative does not achieve the requirement to require callers to present a supported credential before receiving model predictions.
D:A managed online endpoint handles synchronous requests using deployed inference code and compute; it differs from batch processing that writes results for large offline inputs; the actual design criterion is to require callers to present a supported credential before receiving model predictions.
Question 233
An ML service must return predictions synchronously to application requests. The team is required to increase the number of serving replicas to handle sustained concurrent inference load. Which option is the best fit?
- Online deployment instance count
- Online endpoint authentication mode
- Managed online endpoint
- Endpoint request timeout
Explanation
Correct Answer
A:Replica count controls deployed serving capacity within available quotas. It does not alter model accuracy or evaluation-set construction.
Incorrect Answers
B:Endpoint authentication controls who may invoke the service; it does not replace identity permissions needed by scoring code to read external data; the outstanding need is to increase the number of serving replicas to handle sustained concurrent inference load.
C:A managed online endpoint handles synchronous requests using deployed inference code and compute; it differs from batch processing that writes results for large offline inputs; this approach leaves the team unable to increase the number of serving replicas to handle sustained concurrent inference load.
D:Timeout settings bound how long requests can run before failing; extending them may mask slow scoring code and is not a substitute for profiling; the needed capability is to increase the number of serving replicas to handle sustained concurrent inference load.
Question 234
An ML service must return predictions synchronously to application requests. An operator is tasked to tune supported request settings for an inference workload that legitimately needs longer processing time. Which mechanism provides that capability?
- Endpoint traffic allocation
- Online deployment instance count
- Endpoint request timeout
- Managed online endpoint
Explanation
Correct Answer
C:Timeout settings bound how long requests can run before failing. Extending them may mask slow scoring code and is not a substitute for profiling.
Incorrect Answers
A:Traffic weights support canary and blue/green testing among deployments; merely registering another model version does not change live request routing; the needed capability is to tune supported request settings for an inference workload that legitimately needs longer processing time.
B:Replica count controls deployed serving capacity within available quotas; it does not alter model accuracy or evaluation-set construction; the requirement, by contrast, is to tune supported request settings for an inference workload that legitimately needs longer processing time.
D:A managed online endpoint handles synchronous requests using deployed inference code and compute; it differs from batch processing that writes results for large offline inputs; the specified outcome instead requires the team to tune supported request settings for an inference workload that legitimately needs longer processing time.
Question 235
A data-processing team scores millions of stored records overnight rather than serving interactive requests. The immediate design goal is to expose an asynchronous scoring interface suited to large datasets and job-style invocation. Which approach addresses that goal?
- Batch retry policy
- Batch deployment
- Batch input data asset
- Batch endpoint
Explanation
Correct Answer
D:A batch endpoint initiates batch inference jobs and produces output artifacts. It is not the same low-latency request-response path as online inference.
Incorrect Answers
A:Retry settings can reattempt failed units under supported batch execution behavior; persistent schema defects should be fixed rather than retried indefinitely; the specified outcome instead requires the team to expose an asynchronous scoring interface suited to large datasets and job-style invocation.
B:A batch deployment holds execution details for batch scoring; the endpoint interface alone does not define each worker's runtime and parallelism; the implementation under review needs to expose an asynchronous scoring interface suited to large datasets and job-style invocation.
C:A data asset gives batch execution a traceable input reference; it does not imply the files were already loaded into memory or transformed; the requested task remains to expose an asynchronous scoring interface suited to large datasets and job-style invocation.
Question 236
A data-processing team scores millions of stored records overnight rather than serving interactive requests. The team is required to specify the model or component, compute target, and settings used by an invoked batch endpoint. Which option is the best fit?
- Batch job output location
- Batch retry policy
- Batch deployment
- Batch endpoint
Explanation
Correct Answer
C:A batch deployment holds execution details for batch scoring. The endpoint interface alone does not define each worker's runtime and parallelism.
Incorrect Answers
A:Batch execution persists result artifacts rather than returning every prediction inline to the initiating client; the output location must be accessible to the job identity; this alternative does not achieve the requirement to specify the model or component, compute target, and settings used by an invoked batch endpoint.
B:Retry settings can reattempt failed units under supported batch execution behavior; persistent schema defects should be fixed rather than retried indefinitely; the actual design criterion is to specify the model or component, compute target, and settings used by an invoked batch endpoint.
D:A batch endpoint initiates batch inference jobs and produces output artifacts; it is not the same low-latency request-response path as online inference; this approach leaves the team unable to specify the model or component, compute target, and settings used by an invoked batch endpoint.
Question 237
A data-processing team scores millions of stored records overnight rather than serving interactive requests. An operator is tasked to provide a registered collection of files as the input to a batch scoring invocation. Which mechanism provides that capability?
- Batch input data asset
- Batch deployment
- Batch endpoint
- Batch mini-batch tuning
Explanation
Correct Answer
A:A data asset gives batch execution a traceable input reference. It does not imply the files were already loaded into memory or transformed.
Incorrect Answers
B:A batch deployment holds execution details for batch scoring; the endpoint interface alone does not define each worker's runtime and parallelism; the needed capability is to provide a registered collection of files as the input to a batch scoring invocation.
C:A batch endpoint initiates batch inference jobs and produces output artifacts; it is not the same low-latency request-response path as online inference; the requirement, by contrast, is to provide a registered collection of files as the input to a batch scoring invocation.
D:Mini-batch sizing affects worker overhead and resource consumption; it is separate from the number of model versions in a registry; the team would still need to provide a registered collection of files as the input to a batch scoring invocation.
Question 238
A data-processing team scores millions of stored records overnight rather than serving interactive requests. The immediate design goal is to write prediction results to an approved storage path so downstream reporting can consume them. Which approach addresses that goal?
- Batch mini-batch tuning
- Batch endpoint
- Batch input data asset
- Batch deployment
- Batch job output location
Explanation
Correct Answer
E:Batch execution persists result artifacts rather than returning every prediction inline to the initiating client. The output location must be accessible to the job identity.
Incorrect Answers
A:Mini-batch sizing affects worker overhead and resource consumption; it is separate from the number of model versions in a registry; the team would still need to write prediction results to an approved storage path so downstream reporting can consume them.
B:A batch endpoint initiates batch inference jobs and produces output artifacts; it is not the same low-latency request-response path as online inference; the specified outcome instead requires the team to write prediction results to an approved storage path so downstream reporting can consume them.
C:A data asset gives batch execution a traceable input reference; it does not imply the files were already loaded into memory or transformed; the implementation under review needs to write prediction results to an approved storage path so downstream reporting can consume them.
D:A batch deployment holds execution details for batch scoring; the endpoint interface alone does not define each worker's runtime and parallelism; the requested task remains to write prediction results to an approved storage path so downstream reporting can consume them.
Question 239
A data-processing team scores millions of stored records overnight rather than serving interactive requests. The team is required to adjust how many input records or files a worker processes per scoring call for throughput and memory balance. Which option is the best fit?
- Batch job output location
- Batch input data asset
- Batch retry policy
- Batch mini-batch tuning
Explanation
Correct Answer
D:Mini-batch sizing affects worker overhead and resource consumption. It is separate from the number of model versions in a registry.
Incorrect Answers
A:Batch execution persists result artifacts rather than returning every prediction inline to the initiating client; the output location must be accessible to the job identity; the requested task remains to adjust how many input records or files a worker processes per scoring call for throughput and memory balance.
B:A data asset gives batch execution a traceable input reference; it does not imply the files were already loaded into memory or transformed; this alternative does not achieve the requirement to adjust how many input records or files a worker processes per scoring call for throughput and memory balance.
C:Retry settings can reattempt failed units under supported batch execution behavior; persistent schema defects should be fixed rather than retried indefinitely; the actual design criterion is to adjust how many input records or files a worker processes per scoring call for throughput and memory balance.
Question 240
A data-processing team scores millions of stored records overnight rather than serving interactive requests. An operator is tasked to recover transient failures in independent scoring work units without rerunning an entire successful workload. Which mechanism provides that capability?
- Batch endpoint
- Batch job output location
- Batch mini-batch tuning
- Batch retry policy
Explanation
Correct Answer
D:Retry settings can reattempt failed units under supported batch execution behavior. Persistent schema defects should be fixed rather than retried indefinitely.
Incorrect Answers
A:A batch endpoint initiates batch inference jobs and produces output artifacts; it is not the same low-latency request-response path as online inference; the outstanding need is to recover transient failures in independent scoring work units without rerunning an entire successful workload.
B:Batch execution persists result artifacts rather than returning every prediction inline to the initiating client; the output location must be accessible to the job identity; this approach leaves the team unable to recover transient failures in independent scoring work units without rerunning an entire successful workload.
C:Mini-batch sizing affects worker overhead and resource consumption; it is separate from the number of model versions in a registry; the needed capability is to recover transient failures in independent scoring work units without rerunning an entire successful workload.