PostgreSQL services from hyperscalers raise downtime conerns • The Register

Evaluation Current analysis suggests prospects are involved concerning the uptime reliability of hyperscalers’ PostgreSQL situations, giving smaller different distributors a gap to fill the hole.

Analysis agency Foundry found that amongst customers of widespread cloud companies for PostgreSQL, 82 p.c expressed concern about cloud area failures, whereas 21 p.c skilled such failures previously yr.

Within the first six months of 2025, database recognition rating service DB-Engines discovered PostgreSQL the largest climber, with a rating enhance of over 13 factors. It’s fourth general, behind Oracle, MySQL, and Microsoft SQL Server. In 2023, the open supply database grew to become the preferred database for builders, doubtless as a result of companies from the main three cloud distributors made it simply out there to builders.

However final week’s Foundry survey of 212 IT decision-makers throughout enterprises and SaaS companies discovered that 91 p.c of organizations presently utilizing PostgreSQL demand not more than 4 minutes of downtime per 30 days, or round 99.99 p.c uptime, whereas 24 p.c purpose for lower than 30 seconds.

The research was commissioned by distributed PostgreSQL vendor pgEdge. Gartner’s Adam Ronthal, veep and analyst, stated was unclear whether or not there’s a real downside with PostgreSQL service reliability SLAs but. “Actually, inbound calls to Gartner’s inquiry service doesn’t clearly establish this as a problem at this cut-off date.”

Nonetheless, he referred to as on cloud service suppliers to trace and doc uptime and reliability to defuse the “rumors” and establish if there’s genuinely an issue with particular companies.

Microsoft gives Azure Database for PostgreSQL as a completely managed service with single server, versatile server, and hyperscale choices. Each AWS and Google provide primary PostgreSQL companies in addition to Aurora and AlloyDB as choices extra tightly built-in with the underlying cloud infrastructure.

Convey on the options

Any customers discovering that these companies don’t meet their uptime wants have a spread of different PostgreSQL and PostgreSQL-compatible companies to select from, with distributed again ends to reinforce reliability. Examples embrace CockroachDB and YugabyteDB.

In the meantime, PlanetScale, which has made its identify with a distributed MySQL service based mostly on YouTube-developed distributed relational database Vitess, launched a PostgreSQL service earlier this month.

Talking to The Register, CEO Sam Lambert stated that reasonably than compromise compatibility, the PlanetScale service is constructed on open supply PostgreSQL and makes use of the proprietary operator the corporate developed for MySQL/Vitess.

“The operator is the piece of magic that type of is what ensures PlanetScale has such sturdy uptime and reliability: it is what manages all the nodes, the state and the topology. It was actually nearly getting PostgreSQL fitted into that,” he stated.

The present service shouldn’t be sharded, however a sharded service is in growth with prospects, Lambert stated. It stated it might be “take a look at equal” for PostgreSQL, however not 100% appropriate. “If anybody needs to make use of that, they need to migrate to the unsharded service, and we’ll work on getting folks to the sharded model from there.”

He stated YugabyteDB and CockroachDB have been “spectacular” distributed database companies, however claims they don’t provide the identical diploma of PostgreSQL compatibility. “They are not even that appropriate,” he stated.

For instance, the PostgreSQL Compatibility Index scores CockroachDB at about 40 p.c, whereas Yugabyte achieves about 85 p.c compatibility.

Lambert stated PlanetScale’s present PostgreSQL service was 100% appropriate whereas the sharded service would doubtless be 99 p.c appropriate when it arrives.

Convex, which gives an open-source reactive database for internet app builders, has already began to maneuver from AWS Aurora to PlanetScale. In a weblog publish the corporate stated: “For years, we have operated Convex on prime of AWS Aurora. The expertise has been okay. Aurora does what it says on the tin: it is dependable, it scales moderately nicely, and it integrates properly with the remainder of AWS. However ‘okay’ is not the place we wish to cease.”

On the choose initiatives already utilizing the brand new service, “efficiency positive aspects are unimaginable”, the company said.

One other firm trying to reap the benefits of the questions on hyperscalers’ PostgreSQL service reliability, and the compatibility of the distributed techniques, is pgEdge, which guarantees multi-master distributed database constructed on open-source PostgreSQL, permitting customers to have a number of grasp databases unfold throughout totally different areas.

Talking to El Reg co-founder and CEO Phillip Merrick stated the totally PostgreSQL-compatible service gives “fairly excessive ranges of excessive availability”.

As a by-product, the service additionally helps deal with the lag downside for utility customers needing knowledge from a database in numerous elements of the world. “For functions which are getting used globally, whenever you’re accessing web sites or apps which are hosted completely within the US, from the UK and Europe, you are going to see a little bit of a lag, simply because the pace of sunshine is a factor, and it takes some time for these bits to commute throughout the ocean.”

Typical prospects are giant organizations, with a quantity in monetary companies, he stated. “Most of our prospects are monetary companies, authorities, different giant enterprises, the place the appliance merely cannot go down, so there’s virtually zero tolerance for any downtime, and consequently, they want excessive ranges of excessive availability, coupled with the necessity to have the ability to failover for catastrophe restoration safety from one geographical area to a different,” he stated.

Gartner’s Ronthal stated distributed PosgreSQL companies can insulate from the failure of a single cloud area. “The organizations that have a tendency to want this diploma of fault tolerance signify a comparatively small phase of the market. The tradeoff is normally round core options. You get in-built distributed processing and transactional consistency, however perhaps hand over another superior DBMS options,” he stated.

Such distributed companies may be an choice for customers of hyperscalers’ PostgreSQL companies in the event that they implement the identical SQL API, Ronthal stated. “The true query is whether or not these choices present compelling value/efficiency, and whether or not RDS customers want that diploma of excessive availability. Some use circumstances do, some don’t. RDS is a transition stepping stone to extra succesful companies – whether or not they’re tightly built-in with the underlying cloud infrastructure like Aurora or AlloyDB, or third events that provide one thing totally different just like the distributed SQL Database distributors.”

In response to the survey, and excellent questions over the reliability of PostgreSQL companies, AWS stated its PostgreSQL-compatible Amazon Aurora DSQL was designed for as much as 99.999 p.c availability whereas Amazon Aurora for PostgreSQL gives a 99.99 p.c uptime SLA and international replication with cross-Area catastrophe restoration in lower than 1 minute. The spokesperson stated AWS supplied documentation to elucidate what prospects ought to count on by way of downtime throughout upgrades.

Google and Microsoft have been additionally contacted concerning the points raised on this article. ®