Cloud Dataproc, which the search giant launched into open beta on Wednesday, is a new piece of its big data portfolio that's designed to help companies create clusters quickly, manage them easily and turn them off when they're not needed.
Enterprises often struggle with getting the most out of rapidly evolving big data technology, said Holger Mueller, a vice president and principal analyst with Constellation Research.
"It's often not easy for the average enterprise to install and operate," he said. When two open source products need to be combined, "things can get even more complex."
An easy way to implement and operate Hadoop and Spark clusters could be of significant value for enterprises, he added. For Google, meanwhile, Cloud Dataproc will ultimately mean more load, utilization and customers, which creates better economies of scale, Mueller noted.
Cloud Dataproc offers a number of advantages over both traditional, on-premises products and competing cloud services, Google said.
Whereas creating Spark and Hadoop clusters on-premises or through Infrastructure-as-a-Service (IaaS) providers can take anywhere from five to 30 minutes, for instance, Cloud Dataproc clusters take 90 seconds or less on average to start, and the same amount of time to scale or shut down. That, in turn, can mean users have more time to spend working with their data.
"When you do self-managed deployment, either on-premises or in the cloud, you're effectively paying in your own time for your clusters," said Greg DeMichillie, director of product management for the Google Cloud Platform. "What Cloud Dataproc allows you to do is shorten the window of time between when you ask a question and when you get insight."
Pricing is 1 cent per virtual CPU in each cluster per hour, and Cloud Dataproc clusters can include pre-emptible instances that have still lower compute prices, thereby reducing costs further. Whereas many providers round up usage to the nearest hour, Cloud Dataproc uses minute-by-minute billing and a 10-minute-minimum billing period.
Cloud Dataproc also offers built-in integration with Google Cloud Platform services such as BigQuery, Cloud Storage, Cloud Bigtable, Cloud Logging and Cloud Monitoring. Companies can use it to extract, transform and load terabytes of raw log data directly into BigQuery for business reporting, for example.
Because the service is managed, companies can use Spark and Hadoop clusters without the assistance of an administrator or special software, Google said. Rather, they can interact with clusters and Spark or Hadoop jobs through the Google Developers Console, the Google Cloud SDK or the Cloud Dataproc REST API; when they're done with a cluster, they can turn it off and avoid spending money needlessly.
The current implementation of Cloud Dataproc features clusters based on Spark 1.5 and Hadoop 2.7.1.