Google's AlloyDB combines PostgreSQL with GCP open structure

AlloyDB brings speed and more predictable pricing to the table.
Written by Chris Preimesberger, Contributor

Choice is always a good thing. Starting today, enterprises with data-overflow issues -- just about every company -- will have a new cloud-based option for storing their files, data, metadata, and logs.

That choice is provided by Google Cloud Platform, which unveiled a new parallel database, AlloyDB, today at Google's I/O 2022 virtual conference. AlloyDB is built atop the popular open-source PostgreSQL DB that has been a standard choice for developers for more than three decades, so the new database could be both familiar and brand new to a good portion of its users. 

According to Gartner Research, 75% of the world's databases are expected to be run in the cloud this year. AlloyDB is expected to bear much of the data load being created by the exploding e-commerce and social network sectors -- plus all of the new applications and data that will be filling digital coffers in preparation for the coming Web3 environment.

What AlloyDB brings to the market

Two of the most important differentiators AlloyDB brings to the table are speed and more predictable pricing, Google Vice President of Databases Andi Gutmans told ZDNet. 

"We can actually run analytical queries up to 100 times faster than open-source Postgres," Gutmans said. "On the transactional side, based on our benchmarks, we run up to about four times faster than open-source Postgres, and about two times faster than Amazon's equivalent offering (Aurora). 

"We put a lot of focus on really making sure that both the performance on the transactional side and then [its extension] to analytical is the best," Gutmans added. "This means that customers who want to do real-time fraud detection, real-time recommendations, real-time inventory management -- they can actually do a lot of this right in their operational store."

As for pricing, storing data in the cloud has always been an inexact science. Most cloud providers, such as Microsoft Azure, AWS, Dell, Oracle, and others, charge users for data egress, computing, IOPS, and other services. 

So Google decided to take the risk on itself and simplify the pricing model, Gutmans said. No access or IOPS fees will be charged for AlloyDB.

"All along, customers have paid for computing, and for the storage that they're using, they pay for IOPS usage," Gutmans said. "That was actually one of the key pain points we heard from some customers -- that this was comprising about 60% of their bill. When it comes to IOPS, they didn't feel they could actually control it. It's really hard to manage because customers can't actually foretell the cost; it depends on how much data is in memory, how much is in storage, and so on. So we dropped those charges."

"Our goal was really to make this a set-and-forget IT experience -- like autopilot," Gutmans said.

Expected AlloyDB use cases

AlloyDB is aimed at DB admins with code stacks that use a full-featured database offering options such as ACID-based (atomicity, consistency, isolation and durability) compliant transactions and stored procedures. Gutmans told ZDNet that he expects AlloyDB to fit quickly into the market and compete directly with conventional DBs from Oracle, IBM, or Microsoft by delivering the necessary features in a cloud-native package that will be easier to control. CFOs will also like that part. 

Here's how AlloyDB differentiates itself against competitors, according to Gutmans:

  • Unlike Oracle, AlloyDB supports PostgreSQL. Additionally, AlloyDB supports both the Postgres implementation and automation features such as autoscaling. It offers four times faster performance than standard PostgreSQL for transactional workloads, and it achieves high availability SLA up to 99.99%, Gutmans said. 

  • AlloyDB offers more than two times faster processing than Amazon Aurora. Additionally, Gutmans said, AlloyDB has a more customer-friendly pricing model, as noted above. Unlike Amazon Aurora, AlloyDB will not charge for I/O, which can be a major source of unpredictable costs -- up to 60% of the total bill for transactional workloads.

  • AlloyDB is a new option for enterprises seeking to transition off proprietary databases, as well as for PostgreSQL users with demanding, high-end applications. It enables users to cost-effectively modernize their proprietary databases and scale their business-critical workloads, Gutmans said. 

In technical terms, AlloyDB is: an intelligent, database-optimized storage service; an optimized, 100% PostgreSQL-compatible database engine; and a service with built-in autopilot capabilities, including built-in integration with Vertex AI, which allows users to call models directly within a query or transaction. 

AlloyDB also has a built-in columnar accelerator to run analytical queries faster than standard PostgreSQL, Gutmans said. It is a flexible tool for developers because of the combination of PostgreSQL and Google Cloud's open infrastructure. With this, developers will be able to use AlloyDB to quickly build applications and scale them -- all with their existing open-source skills, Gutmans said.

Editorial standards