How to improve Vertex AI Feature Store serving performance during ingestion?

You recently deployed a model to a Vertex AI endpoint and set up online serving in Vertex AI Feature Store. You have configured a daily batch ingestion job to update your featurestore. During the batch ingestion jobs, you discover that CPU utilization is high in your featurestore’s online serving nodes and that feature retrieval latency is high. You need to improve online serving performance during the daily batch ingestion. What should you do?

Schedule an increase in the number of online serving nodes in your featurestore prior to the batch ingestion jobs
Enable autoscaling of the online serving nodes in your featurestore Source Reference Answer
Enable autoscaling for the prediction nodes of your DeployedModel in the Vertex AI endpoint
Increase the worker_count in the ImportFeatureValues request of your batch ingestion job

Community Votes

64%

36%

64% of anonymous learners picked answer B. Votes are pick records left by other test-takers — they are not the verified answer.

Community Insight

This question tests the ability to choose between manual scaling and autoscaling for Vertex AI Feature Store nodes to handle ingestion-related traffic spikes.

To address high CPU and latency in Vertex AI Feature Store during batch ingestion, the consensus is to enable autoscaling for online serving nodes. This allows the infrastructure to dynamically adapt to the increased load without manual intervention.

Selecting Option A (scheduling node increases) is a common error because users assume manual provisioning is more reliable than autoscaling for predictable jobs.

Community Discussion (8 comments)

mouthwash 👍 1 Selected: B

Answer is B, coz how do you predict there will be a problem before the batch ingestion job? Seems preemptive and may be unnecessary. B aligns more coz error has happened and now you enabling autoscaling, so that in the future it will autoscale.

tardigradum 👍 1 Selected: A

Agree with bobjr

Prakzz 👍 1 Selected: A

https://cloud.google.com/vertex-ai/docs/featurestore/managing-featurestores Specifically mentioned here that --> If CPU utilization is consistently high, consider increasing the number of online serving nodes for your featurestore.

bobjr 👍 2 Selected: A

Gemini + Perplexity ai + ChatGPT votes A Because : B. Enable Autoscaling: While autoscaling can be useful, it might not react quickly enough to sudden spikes in traffic during batch ingestion. Scheduling the increase ensures that the resources are available when needed.

cruise93 👍 1 Selected: D

This question is valid for the Legacy feature store. https://cloud.google.com/vertex-ai/docs/featurestore/ingesting-batch#import_job_performance

daidai75 👍 2 Selected: B

https://cloud.google.com/vertex-ai/docs/featurestore/managing-featurestores?&_gl=1sswg5e_gaNDE2OTc3OTAzLjE3MDU4OTQ5OTE._ga_WH2QY8WWF5*MTcwNTkzNDM0NS40LjAuMTcwNTkzNDM0NS4wLjAuMA..&_ga=2.242492743.-416977903.1705894991#online_serving_nodes

b1a8fae 👍 2 Selected: B

Vertex AI Feature Store provides two options for online serving: Bigtable and optimized online serving. Both options support autoscaling, which means that the number of online serving nodes can automatically adjust to the traffic demand. By enabling autoscaling, you can improve the online serving performance and reduce the feature retrieval latency during the daily batch ingestion. Autoscaling also helps you optimize the cost and resource utilization of your featurestore.

pikachu007 👍 2 Selected: B

Option A: Manually scheduling node increases requires prior knowledge of batch ingestion times and might not be as responsive to unexpected workload spikes. Option C: Autoscaling prediction nodes in the Vertex AI endpoint might help with model prediction latency but doesn't directly address feature retrieval latency from the featurestore. Option D: Increasing worker_count in the batch ingestion job could speed up ingestion but might further strain online serving nodes, potentially worsening latency.

Comments & Corrections

No comments yet — spotted an error or have a note? Share it below.

Expert Analysis

Why the Answer Is Correct

Enabling autoscaling (Option B) allows the online serving nodes to automatically scale out in response to the increased load caused by the daily batch ingestion job. This directly addresses the high CPU utilization and feature retrieval latency by ensuring sufficient resources are available when needed, aligning with Google Cloud's best practices for managing Feature Stores.

Why the Other Options Are Wrong

Option A suggests manually scheduling node increases, which is less flexible and requires precise prediction of the load. Option C is incorrect because it targets the Vertex AI endpoint's prediction nodes rather than the Feature Store's serving nodes where the bottleneck exists. Option D focuses on increasing the ingestion worker count, which speeds up data import but does not resolve the serving performance issues on the online nodes.

Community Comment Notes

There is a notable split in the community between Options A and B. Proponents of Option A argue that manual scheduling ensures resources are ready before the spike starts, avoiding potential autoscaling lag. However, the majority and official documentation references favor Option B for its ability to handle workload dynamics automatically, as noted in comments citing Google Cloud docs.

Official Reference

https://cloud.google.com/vertex-ai/docs/featurestore/managing-featurestores

Exam Strategy

Focus on identifying the specific component causing the bottleneck—here, it is the Feature Store serving nodes, not the prediction endpoint. Prefer autoscaling over manual scheduling for handling recurring but variable workloads in cloud environments.

Related Analysis

Practice All PMLE Questions

Access 65 questions with complete answers and detailed explanations.

View Full PMLE Practice Test →

← Back to PMLE Study Guide