Google announces AlloyDB, a faster, hosted version of PostgreSQL

Data-laden users will have a new option for storing their information in the cloud now. Google Cloud Platform (GCP) today announced a new database option called AlloyDB that’s built around the PostgreSQL open-source database that has been a popular choice for developers for more than 30 years. 

The new database is designed to appeal to users with a code stack that relies upon a full-featured database offering options like atomicity, consistency, isolation and durability (ACID)-compliant transactions, stored procedures or triggers. Google’s team believes that it will compete directly with legacy offerings from companies like Oracle, IBM or Microsoft by delivering the classic features in a modern, cloud-native package. 

“We’ve got lots of customers, like travel agencies, retailers, auto manufacturers, or financial services who bought these very expensive, proprietary databases and are trying to really break free from them and move on to open source.” explained Andi Gutmans, the general manager and vice president of databases at Google Cloud.

PostreSQL in the cloud

The PostgreSQL platform is a popular option because of its strong performance, a broad set of features and a large community of developers. The open-source license is attractive because users feel less locked into one company. If the contract terms or the price grows too onerous, they can move to another service provider supporting the product or build up a team in-house. 

“You can kind of guess why customers want to move,” said Gutmans. “The cost is definitely one part, but there are a lot of prohibitive licensing terms. They have audits being done against them. There are a lot of, I would say, unfriendly practices.” 

Google’s new version will be hosted in their cloud and priced as a service. The new pricing model is designed to be simpler and free from the kind of hidden charges that often create large and unexpected bills. Their model, for example, won’t charge for I/O, a common extra in some other contracts from cloud providers like Amazon Web Services (AWS). 

The new offering from Google is joining several other companies that are building database products around PostgreSQL. Some cloud providers like DigitalOcean, Vultr and AWS are rolling out managed versions of popular open-source databases like MySQL, Redis and PostgreSQL. These products deliver the stock version of the database while handling many of the chores of installing the software, configuring the server and keeping it up-to-date as new security patches appear. 

Meanwhile, other companies are building more elaborate versions around the open-source database, while adding some new features that allow them to create a new brand. Companies like Yugabyte and Fly.io are creating versions of PostgreSQL that scale to support large datasets distributed around the world. They manage many of the chores of synchronizing the data between different instances and shards. Some like Oracle and PlanetScale are doing something similar with MySQL, another popular open-source option. 

Google plans to distinguish itself with faster performance and a rock-solid service-level agreement. They’ve rewritten some core storage routines to speed up both transactional and analytical queries. Their initial internal benchmarks suggest that their version will be four times faster than stock PostgreSQL and twice as fast as Amazon’s Aurora, another competitor following a similar path. 

They’ve also included a columnar accelerator that stores the dataset in columns, an approach that can speed up complex searches and queries. Analytical tasks like creating reports or watching for important anomalies often run more efficiently in environments where the data is stored in columns, while transactions can be faster when the data is stored in rows. AlloyDB will offer users the ability to configure the storage to support their pattern of usage. 

“Customers want to analyze the data in real time in order to make additional decisions very, very quickly about things like personalization or fraud detection or so forth,” explained Gutmans. “What we’ve actually added Is an analytical capability, so now you have a hybrid transactional and analytical system. They can run analytics up to 100 times faster than [stock] open-source PostgreSQL.”

Integrations and compatibility

Google also integrated the database with Vertex AI, one of its options for building and deploying machine learning models. Developers will be able to work directly with these models with queries and stored procedures. 

PostgreSQL is also popular because the community has created a number of extensions that add features for particular applications. Mapmakers and developers working with location data, for example, rely upon PostGIS, a version optimized for storing and searching collections of points specified by latitudes and longitudes. 

“[Compatibility] was actually a core design principle for us.” said Gutmans. “This is why we decided to take PostgreSQL and extend it as opposed to building a PostgreSQL compatible system. We have a PostgreSQL API on spanner, right? But you can’t build that to be 100% compatible, and so what we’ve done here is we’ve stayed true to PostgreSQL.” 

Gutmans estimates that AlloyDB will begin with support for more than 50 of the most popular extensions and add new ones following customer demand. 

What execs are saying so far

“AlloyDB is fully compatible with PostgreSQL and can transparently extend column-oriented processing.” said Takuya Ogawa, a lead product engineer at Plaid who is testing a pre-released version of Google’s revamped PostgreSQL. 

“We think it’s a new, powerful option with a unique technical approach that enables system designs to integrate isolated OLTP, OLAP and HTAP workloads with minimal investment in new expertise.” 

Others agreed and emphasized the combination of full compatibility with cloud availability. 

“AlloyDB provides us with a compelling relational database option with full PostgreSQL compatibility, great performance, availability and cloud integration. We are really excited to co-innovate with Google and can now benefit from enterprise-grade features while cost-effectively modernizing from legacy, proprietary databases.” said Bala Natrajan, senior director of data infrastructure and cloud engineering at PayPal.

Some execs are attracted to the support and management that Google offers. The sales literature emphasizes the service’s ability to handle many of the scaling, back up and replication tasks. Google will deploy machine learning-based models to learn from users and adapt to their query patterns. 

“With AlloyDB, we have significantly increased throughput, with no application changes to our PostgreSQL workloads. And since it’s a managed service, our teams can spend less time on database operations and more time on value-added tasks,” said Sofian Hadiwijaya, CTO and cofounder of Warung Pintar.

Google believes that customers like this will form the basis for a strong customer base. They need compatibility with their current legacy options at a lower price with open-source software’s flexibility. 

“The predictions are that 70% of new in-house applications will be developed on open source and 50% of existing proprietary databases will either have migrated to open source or begin the process of converting,” said Gutmans. “This is just something we’ve been hearing time and time again from our customers.”