Google Cloud today introduced "a big change in BigQuery," including new pricing tiers, autoscaling capabilities, and a compressed storage billing model for the serverless data warehouse, Gerrit Kazmaier, GM and VP for databases, data analytics, and Looker at Google Cloud, told a group of reporters ahead of the provider's Data Cloud and AI [artificial intelligence] Summit.

The serverless nature of BigQuery means customers can autoscale unpredictable workloads by adding small compute increments based on real-time needs. "Basically, a workload gets exactly what it needs for the time it needs," Kazmaier explained. This serverless compute architecture, he added, paved the way for BigQuery's new pricing model based on autoscaling.

BigQuery editions, also announced today, is designed to improve the warehouse's flexibility and affordability. Editions is a tiered model that offers different sets of features and capabilities handpicked by customers on a yearly basis. Customers can mix and match from the Standard, Enterprise, and Enterprise Plus editions based on their desired price and performance for each workload.

"In this year of both challenges and opportunities, customers need a flexible data cloud which allows them to optimize and to have the right financial governance in place," Kazmaier said.

The alternative, Kazmaier explained, is a virtual machine (VM)-based offering that typically charges for use of a full data warehouse with a fixed capacity. "With VMs, you need to start them and stop them, and they always carry a degree of over or under provisioning in them because they are such coarse-grained units of scale," he said.

BigQuery's serverless architecture, however, helps customers avoid paying for unused capacity by provisioning capacity as it's needed. "We constantly shuffle our feed around so that our queries are running optimally. And this was not happening on a VM-based case, but on fine granular compute units," he said.

And for BigQuery editions customers, Google Cloud introduced a compressed storage billing model designed to lower costs based on the amount of structured and unstructured data stored in the serverless warehouse. "It is basically giving our customers access to data in a highly compressed format using a proprietary multistage compression process, and [they] only pay for the data that is being physically stored," Kazmaier said.

Exabeam, for example, is using the BigQuery editions compression model to achieve a 12 to 1 compression rate, which is allowing the company to grow its data footprint while paying less to do so, he added.

While this new model is targeted at customers seeking to optimize for "spiky" or "exploratory" data workloads, it also significantly impacts stable workloads. "The most interesting thing is that when we mastered the new autoscale mode and the new way of scheduling capacity, we also saw up to a 40% efficiency increase even for stable workloads," Kazmaier explained.