Why Google Cloud is battling AWS and Azure in the red-hot PostgreSQL market
- 12 May, 2022 07:45
Thomas Kurian (CEO - Google Cloud)
With more enterprises shifting their data from legacy to open source databases in the cloud, Google Cloud Platform (GCP) is offering a PostgreSQL-compatible, fully managed database-as-a-service (DBaaS), dubbed AlloyDB, now in public preview and intended to take on the likes of Amazon Aurora and Microsoft Azure's Database for PostgreSQL.
By 2022, 75 per cent of all databases will be deployed or migrated to a cloud platform, with only five per cent ever considered for repatriation to on-premises systems, according to Gartner. The trend is fuelled by the move on the part of enterprises to use databases for analytics, Gartner said.
In addition, Gartner outlined, more than 50 per cent of legacy databases are in transition to open source and more than 70 per cent of new application development is happening on open source databases. Meanwhile, many enterprises are choosing PostgreSQL for their database management system.
Why are enterprises scooping up PostgreSQL?
Google Cloud, AWS and Microsoft have moved their focus to PostgreSQL because of its rising popularity among other databases, analysts added.
"Postgres is inarguably in the midst of a resurgence in interest and usage. This is in large part because it is an open source database with many potential providers, and because it is a versatile database that has increasingly been adapted for all manner of workloads," said Stephen O'Grady, principal analyst at RedMonk.
In addition, said Tony Baer, principal and founder of dbInsight, there is also a huge ecosystem of PostgreSQL-based databases leveraging the core technology and skills base that is impossible to ignore, positioning it as a default, standard enterprise-grade open source database.
Other analysts believe that though PostgreSQL is yet to overtake the popularity of MySQL, from Oracle, it has advantages over it.
"When comparing MySQL and PostgreSQL, organisations prefer PostgreSQL's transactional and analytical capabilities, extended support for spatial data, broader SQL support, enhanced security and governance, and expanded support for programming languages, driving more substantial growth," said Noel Yuhanna, principal analyst at Forrester.
According to data from relational databases knowledge platform db-engines.com, PostgreSQL has been steadily rising in popularity and is currently the fourth most-popular RDBMS (relational database management system) and fourth most-popular product cited among all databases in their rankings.
The constant rise in popularity has forced hyperscalers to create database services built on PostgreSQL, said Doug Henschen, principal analyst at Constellation Research.
"In most cases the cloud vendors are pouring resources into offering migration services that make it easier to migrate workloads onto these Postgre-based services, and they're also upping compatibility features with Oracle Database in particular. The availability of multiple services and improvements in compatibility with commercial databases has only encouraged more migrations and increased popularity for PostgreSQL," Henschen said.
Google has recently added an Oracle-to-PostgreSQL schema conversion and data replication capabilities to its Database Migration Service.
Here's the difference between Google's AlloyDB and CloudSQL
Google claims that its new AlloyDB, as a PostgreSQL managed DBaaS, is for "high-end databases" in contrast to its existing CloudSQL PostgreSQL managed database service, which the company claims is also seeing "good" traction.
"We saw that there were certain gaps that we could and needed to fill, to truly make sure that the highest end databases could actually move over. And these are gaps in their in areas of scale, availability, extending the use cases a bit beyond transactional to also support analytical workloads," said Andi Gutmans, general manager of databases at Google Cloud.
AlloyDB leverages Google Cloud's infrastructure abilities in the area of storage, compute and networking while combining it with Postgres compatibility, Gutmans said, adding that it has faster performance and scalability that the existing CloudSQL for PostgreSQL service.
Storage layer is key to PostgreSQL performance
"We find that the key ingredient is the new storage layer that boosts the performance of PostgreSQL," said Forrester's Yuhanna.
"Unlike MySQL, which supports multiple storage engines, PostgreSQL traditionally has only relied on a single storage engine. Although AlloyDB leverages a proprietary cloud-native storage layer, it offers a 100 per cent PostgreSQL compatibility, allowing existing PostgreSQL applications to migrate without application or database changes," Yuhanna said, adding that he expects to see strong traction for the new service as enterprises expand their PostgreSQL usage.
Google has also added accelerators along with AI and machine learning capabilities to the managed DBaaS to ensure ease of management with better performance, Gutmans said.
"We use machine learning and algorithms to basically take away a lot of the manageability issues that customers have had in the past with things like Postgres and do it for them. These are things like auto-VACUUM management, how we configured the instance for best performance, configuring the storage and memory management, etc.," Gutmans said. Auto-VACUUM refers to a process that cleans up redundant data.
AlloyDB also comes with an integration of the Vertex AI platform, a workbench designed for data scientists working on machine learning projects, and according to Gutmans enterprises could "actually do real time inferencing as part of transactions."
An example of such real-time inferencing, which Gartner calls augmented transactions, is to activate or call a fraud detection module within a transaction for a credit card entry for expenditure.
In fact a large number of cloud database services are beginning to take advantage of machine-learning to do continuous, closed-loop analysis for self-tuning and optimisation, Constellation's Henschen said, adding that the simplification of maintenance through self-tuning and optimisation is not unique.
"But in AlloyDB's case, it optimises both row-based and columnar storage simultaneously, so it can serve both transactional and analytical needs and does not have to be tuned, or so Google claims, for one need or the other," Henschen said.
"This 'trans-analytical' performance also supports the integration with Vertex AI. The benefit is high-performance in scenarios such as lending, risk, churn-analysis and so on where you're trying to do business with customers in real-time while simultaneously supporting deep, data-science-backed analysis."
Postgres-compatible v. PostgreSQL
Google Cloud's new AlloyDB for PostgreSQL is a Postgres-compatible service, which is different than a PostgreSQL database service like GCP's CloudSQL.
This means that it is built to be different (faster, more manageable, reliable and scalable) than the standard PostgreSQL RDMS while taking the basic fundamentals and semantics from it, analysts said.
"There are many PostgreSQL-compatible database that are not PostgresSQL, but they use all the SQL semantics of that database to take advantage of the popularity and familiarity of how you program/query PostgreSQL," Henschen said.
It also means that developers can build or shift applications built on PostgreSQL onto this database. The new service, which the company claims is "100 per cent Postgres compatible," supports more than 50 extensions from the Postgres ecosystem, said Google's Gutmans.
Pricing and competition in the PostgreSQL market
Google claims that the new managed DBaaS offers better price performance than rivals because it has tweaked its pricing strategy to exclude onerous input/output charges that most enterprises have to bear for managed database services.
Enterprises will be charged on compute, storage and other services they consume, Gutmans said, adding that this will make it easier for enterprises to budget their expenditure.
But AlloyDB for PostgresSQL service will have heavy competition from the likes of Amazon Aurora, Oracle Autonomous Databases and Oracle mySQL HeatWave, both from a speed/cost point of view, as well as for OLAP (online analytical processing) and OLTP (online transaction processing.
"I don't see Aurora as addressing analytical requirements as extensively as AlloyDB, though it does integrate with SageMaker to support the same sort of real-time inferencing scenarios," Henschen said.
"Meanwhile, Oracle is well known for promoting Oracle Database as a 'universal database' that can do it all, but it's notable that it has separate Oracle Autonomous Database services that are tuned for either transactional or data warehouse/analytical use," Henschen said, noting that there are no real world deployments of AlloyDB yet.
Google's move to launch the new managed DBaaS could be seen as a strategy to cover as much market as possible, RedMonk's Grady said.
"Much as AWS offers both vanilla, standard Postgres and a proprietary, Postgres-compatible alternative in Aurora, Google appears to be catering to both markets that value the standard of an open source database and those willing to sacrifice the all open source base in return for a Postgres-compatible higher performing database," Grady said.
Amazon Aurora supports both MySQL and PostgreSQL databases.