Google Scales Cloud Bigtable NoSQL Database

Google today announced a series of generally available feature updates for its cloud database service Bigtable, designed to help improve scalability and performance.

Google Cloud Bigtable is a managed NoSQL database service that can handle both analytical and operational workloads. Among the new updates Google is bringing to Bigtable is increased storage capacity, with up to 5TB of storage now available per node, an increase from the previous 2.5TB limit. also enhanced autoscaling capabilities, so a DB cluster will automatically grow or shrink as needed based on demand. The Bigtable update is complemented by better visibility into database workloads, with the goal of enabling better troubleshooting of issues.

The new features announced in Bigtable demonstrate a continued focus on increased automation and augmentation that are becoming table stakes for modern cloud services.

Adam RontalAnalyst, Gartner

“The new features announced in Bigtable demonstrate a continued focus on increased automation and augmentation that are becoming table stakes for modern cloud services,” said Adam Ronthal, analyst at Gartner. “They also pursue the goal of improving price and performance – which is quickly becoming the key metric for evaluating and managing any cloud service – and observability, which serves as the basis for improved governance and optimization. financial.”

How Autoscaling Changes Google Cloud Bigtable Database Operations

One of the promises of the cloud has long been the ability to elastically scale resources as needed, without requiring new physical infrastructure for end users.

Programmatic scaling has always been available in Bigtable, according to Anton Gething, Bigtable Product Manager at Google. He added that many Google customers have developed their own autoscaling approaches for Bigtable through programmatic APIs. Spotify, for example, has made available an open source implementation of Cloud Bigtable autoscaling.

“TodayThe Bigtable release introduces a native autoscaling solution,” Gething said.

He added that native autoscaling directly monitors Bigtable servers, to be very responsive. As a result, demand changes, and so does the size of a Bigtable deployment.

The size of each Bigtable node is also increased in the new update. Previously, Bigtable had a maximum storage capacity of 2.5 TB per node; which is now doubled to 5TB.

Gething said users don’t have to upgrade their existing deployment to take advantage of the increased storage capacity. He added that Bigtable has a separation of compute and storage, allowing each type of resource to scale independently.

“This storage capacity update is intended to provide cost optimization for storage-centric workloads that require more storage without the need to increase compute,” Gething said.

Google is bringing native autoscaling to its managed NoSQL database, allowing deployments to auto scale or grow automatically as needed.

Optimize Google Cloud Bigtable database workloads

Another new feature that has landed in Bigtable is a feature known as cluster group routing.

Gething explained that in a replicated Cloud Bigtable instance, cluster groups provide finer control over high availability deployments and better workload management. Prior to the new update, he noted that a user of a replicated Bigtable instance could route traffic to any of his Bigtable clusters in single-cluster routing mode, or to all of his clusters in single-cluster routing mode. multi-cluster routing. He said cluster groups now allow customers to route traffic to a subset of their clusters.

Google has also added a new CPU usage metric per application profile, which allows greater visibility into the performance of a given application workload. Although Google provided some visibility into CPU usage to Bigtable users before the new update, Gething explained that the new update provides new dimensions of visibility into the methods of accessing data queries. and on the database tables consulted.

“Before these extra dimensions, troubleshooting could be difficult,” Gething said. “You would have visibility into the cluster’s CPU usage, but you wouldn’t know what application profile traffic was using the CPU, or which table was being accessed with which method.”

Maria H. Underwood