Deal of The Day! Hurry Up, Grab the Special Discount - Save 25% - Ends In 00:00:00 Coupon code: SAVE25
Welcome to Pass4Success

- Free Preparation Discussions
Mail Us support@pass4success.com
Location US
Mob_nav_barsMENU

Google Professional Machine Learning Engineer Exam - Topic 6 Question 71 Discussion

Actual exam question for Google's Professional Machine Learning Engineer exam
Question #: 71
Topic #: 6
[All Professional Machine Learning Engineer Questions]

You recently deployed a model lo a Vertex Al endpoint and set up online serving in Vertex Al 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 featurestores 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?

Show Suggested Answer Hide Answer
Suggested Answer: 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.Reference:

Online serving | Vertex AI | Google Cloud

New Vertex AI Feature Store: BigQuery-Powered, GenAI-Ready | Google Cloud Blog


by Cecilia at Feb 11, 2024, 04:02 PM

Limited Time Offer

25%

Off

Get Premium Professional Machine Learning Engineer Questions as Interactive Web-Based Practice Test or PDF

Contribute your Thoughts:

0/2000 characters
Submit Cancel
German
4 months ago
Definitely go with B, it’s the most efficient way to manage resources!
upvoted 0 times
...
Jin
4 months ago
Wait, can increasing worker counts really fix latency issues?
upvoted 0 times
...
Demetra
4 months ago
C seems off, it's about the feature store, not the model.
upvoted 0 times
...
Kaitlyn
4 months ago
I think B is better, autoscaling can really help with spikes.
upvoted 0 times
...
Jacklyn
4 months ago
A is a solid choice for handling high CPU usage.
upvoted 0 times
...
Evan
5 months ago
I wonder if increasing the worker counts in the batch job, as in option D, would actually help with the online serving performance during ingestion? It seems a bit off to me.
upvoted 0 times
...
Detra
5 months ago
I feel like we practiced a similar question where autoscaling was a key factor, so option B seems familiar and might be the best approach here.
upvoted 0 times
...
Angella
5 months ago
I'm not entirely sure, but I think increasing the number of online serving nodes beforehand, like in option A, could also help with the latency issue.
upvoted 0 times
...
Junita
5 months ago
I remember we discussed the importance of scaling during high-load periods, so maybe option B about enabling autoscaling is the right choice?
upvoted 0 times
...
Lilli
5 months ago
This seems like a pretty straightforward question. The problem is clearly with the feature store's online serving performance during the batch ingestion job, so the solution should be focused on that. Option B, enabling autoscaling of the online serving nodes, seems like the best choice here. That way, the feature store can automatically scale up to handle the increased load without any manual intervention. I'm confident that's the right answer.
upvoted 0 times
...
Gretchen
5 months ago
Okay, I think I've got this. The key here is that the issue is specifically with the feature store's online serving performance during the batch ingestion job. So the solution should be focused on the feature store, not the deployed model. That rules out option C. Between A and B, I'd go with B and enable autoscaling of the online serving nodes. That way, the feature store can automatically scale up to handle the increased load during the batch ingestion process.
upvoted 0 times
...
Emeline
5 months ago
This seems like a straightforward question about optimizing the performance of a Vertex AI feature store during batch ingestion. I think the key is to identify the bottleneck, which in this case is the high CPU utilization and latency in the online serving nodes. Based on that, I would go with option B and enable autoscaling of the online serving nodes. That should automatically scale up the resources to handle the increased load during batch ingestion.
upvoted 0 times
...
Jeffrey
5 months ago
Hmm, I'm a bit unsure about this one. The question mentions high CPU utilization and latency in the online serving nodes, but it's not clear if the issue is with the feature store itself or the deployed model. I'm leaning towards option B to enable autoscaling of the online serving nodes, but I'm also wondering if option C to enable autoscaling for the prediction nodes might be a better solution. I'll need to think this through a bit more.
upvoted 0 times
...
Oretha
5 months ago
Okay, let's see here. The first query with the LIKE operator could be selective, but the '!-NULL' part seems a bit odd. I'll need to double-check the syntax on that one.
upvoted 0 times
...
Evangelina
6 months ago
Using $HTTP_POST_FILES instead of $_FILES? That seems a bit outdated. I'll stick with the recommended $_FILES approach.
upvoted 0 times
...
Tula
6 months ago
Okay, I've got this. The question is asking about how to get host data into Firepower, so the answer is clearly option A - configuring a Network Discovery policy. That's the policy that's responsible for discovering and monitoring the assets on the network.
upvoted 0 times
...
Asuncion
6 months ago
Hmm, this is a bit complex. I'll need to refer back to my QoS notes and make sure I'm considering all the key details here. Gotta be careful not to miss anything.
upvoted 0 times
...
Sanjuana
2 years ago
I think scheduling an increase in the number of online serving nodes could also help.
upvoted 0 times
...
Slyvia
2 years ago
I disagree. I believe increasing the worker counts in the batch ingestion job would be more effective.
upvoted 0 times
...
Isadora
2 years ago
That sounds like a good idea. It could help improve performance during the batch ingestion.
upvoted 0 times
...
Larue
2 years ago
I think we should enable autoscaling of the online serving nodes.
upvoted 0 times
...
Jina
2 years ago
But what about option D? Increasing the worker counts in the batch ingestion job could also help distribute the load and reduce the impact on online serving, no?
upvoted 0 times
Walton
2 years ago
D) Increasing the worker counts in the batch ingestion job could also help distribute the load and reduce the impact on online serving.
upvoted 0 times
...
Vi
2 years ago
A) Schedule an increase in the number of online serving nodes in your featurestore prior to the batch ingestion jobs.
upvoted 0 times
...
Percy
2 years ago
D) Increase the worker counts in the importFeaturevalues request of your batch ingestion job.
upvoted 0 times
...
Shelba
2 years ago
C) Enable autoscaling for the prediction nodes of your DeployedModel in the Vertex AI endpoint.
upvoted 0 times
...
Odette
2 years ago
B) Enable autoscaling of the online serving nodes in your featurestore
upvoted 0 times
...
Mitzie
2 years ago
A) Schedule an increase in the number of online serving nodes in your featurestore prior to the batch ingestion jobs.
upvoted 0 times
...
...
Maia
2 years ago
I'm leaning towards option B, enabling autoscaling of the online serving nodes. That should help the featurestore handle the increased load without us having to manually adjust the node count.
upvoted 0 times
...
Sherly
2 years ago
Ha, imagine if we had to manually adjust the worker counts every time. 'Okay, everyone, stop what you're doing, it's time for the daily batch ingestion! Quick, someone count the nodes and tell me how many workers we need!'
upvoted 0 times
...
Corrinne
2 years ago
Hmm, I'm not sure. Autoscaling seems like the more elegant solution to me. I don't want to have to manually adjust the worker counts every time we have a batch ingestion job.
upvoted 0 times
...

Save Cancel