Migrate on-prem Hadoop to a managed, fault-tolerant, cost-effective cloud solution for long-running batch jobs. What should you deploy?

This option is correct because Dataproc is Google Cloud's managed Hadoop and Spark service, making it ideal for migrating on-prem Hadoop. Using standard persistent disks is cost-effective for long-running batch jobs where I/O performance isn't the absolute bottleneck. Incorporating 50% preemptible workers further reduces costs while maintaining fault tolerance, as Dataproc can gracefully handle preemption. Storing data in Cloud Storage (GCS) decouples storage from compute, offering high availability, durability, and scalability, and the gs:// prefix is the standard way to access GCS from Dataproc. The incorrect options are: Using SSD persistent disks is more expensive and generally not necessary for most long-running batch jobs unless very high I/O is critical. Installing Hadoop/Spark on Compute Engine instances requires manual management, which defeats the purpose of a managed solution and increases operational overhead. Storing data in HDFS on preemptible Compute Engine instances is highly unreliable and not fault-tolerant, as data would be lost upon preemption, making it unsuitable for long-running jobs.