Google Professional Machine Learning Engineer Question 399
Single answerGoogle Cloud PlatformYour company has deployed a machine learning model for real-time fraud detection on Google Cloud. Due to regulatory and compliance requirements, part of the data must be processed on-premises, while other parts of the data can be processed in Google Cloud. Additionally, your organization is exploring the possibility of using another cloud provider for backup and redundancy. Which approach best supports this hybrid and multicloud strategy while ensuring performance and scalability for the machine learning system?
- A
Use Anthos to manage ML workflows across on-premises and multiple clouds, enabling a unified and consistent deployment strategy.
- B
Migrate all data to Google Cloud and use Vertex AI for model training and serving, ensuring all workloads run in a single environment.
- C
Deploy independent ML pipelines on each environment (on-premises, Google Cloud, and the other cloud provider) and manually synchronize results.
- D
Use TensorFlow Extended (TFX) to build a hybrid pipeline capable of running seamlessly across on-premises and cloud environments without additional management layers.
Show answer and explanation
Correct answer: A
Explanation
Anthos is a hybrid and multicloud platform that allows organizations to consistently manage workloads across on-premises and multiple cloud providers. It supports deploying and managing containers, enabling a unified approach to ML workflows while addressing compliance, performance, and scalability needs. This makes it the most suitable option for the given scenario.
- A. Correct.
Anthos provides a consistent platform for managing hybrid and multicloud environments, making it the best choice for ensuring unified management, scalability, and compliance across on-premises and multiple clouds.
- B. Incorrect.
Migrating all data to Google Cloud does not comply with the regulatory requirements to process part of the data on-premises. It also doesn't address the multicloud need for redundancy.
- C. Incorrect.
Deploying independent ML pipelines increases complexity and introduces challenges with synchronization, consistency, and scalability, making it an inefficient approach.
- D. Incorrect.
While TensorFlow Extended (TFX) is useful for building ML pipelines, it doesn't offer the management and orchestration capabilities needed for hybrid and multicloud strategies.