Q: 7
Your data science team has requested a system that supports scheduled model retraining, Docker
containers, and a service that supports autoscaling and monitoring for online prediction requests.
Which platform components should you choose for this system?
Options
Discussion
B. not A. App Engine doesn't fully support containers or ML monitoring, pretty sure B is the right stack here.
Makes sense to go with B since only Vertex AI Prediction supports autoscale and Docker containers together.
B . Vertex AI Pipelines does scheduled retraining, Vertex AI Prediction covers online predictions with autoscaling, and Model Monitoring checks the deployed models. Not totally sure but pretty sure the others miss something here. Can someone confirm?
Yeah, definitely B for this one
Its B here. Vertex AI Pipelines can handle scheduled retraining, and Vertex AI Prediction supports custom containers plus autoscaling for online prediction. Model Monitoring completes the stack. Only doubt is if Airflow is a strict requirement, but I think B covers all asks.
C or D? Cloud Composer handles scheduling, and Vertex AI Prediction covers serving, so both sound plausible to me. Not sure if BigQuery ML's lack of Docker support is a dealbreaker here though.
I don’t think D fits the requirements, since App Engine won’t cover serving models in Docker for online prediction. B is the only option with Vertex AI Prediction and Model Monitoring, so it checks all the boxes here (retraining, Docker support, autoscale, monitoring). If I missed something about Composer let me know, but pretty confident on B.
Model monitoring and autoscaling only really fit with B here. The nitpick for me: if online prediction didn't require custom containers, C could be tempting, but the Docker part rules it out. Pretty sure it's B, but happy to be corrected if anyone's seen Composer pull this off lately.
Option B is correct. Only Vertex AI Prediction supports deploying Docker containers with autoscaling, plus Model Monitoring handles observability. C's tempting, but BigQuery ML can't run custom containers, so it's out. Anyone see a use case where D would fit better?
Yeah, it's gotta be B for this one
Be respectful. No spam.