Google Cloud Dataproc: Making Spark and Hadoop Easier, Faster, and Cheaper

- Per minute billing

- Industry-leading startup times (get a cluster up in under 90 seconds)

- Fully integration with BigQuery, Bigtable, Cloud Logging, etc.

- Use Preemptible VMs

- Use GCS as your storage engine instead of HDFS (separation of storage and compute, easy multitenancy/concurrency)