PlanetScale Extends Database Platform to PostgreSQL
Briefly

PlanetScale Extends Database Platform to PostgreSQL
"PlanetScale has announced the general availability of its managed sharded Postgres service, built for performance and reliability on AWS or Google Cloud. PlanetScale Metal is built on local NVMe drives and provides "Unlimited I/O," with customers typically running out of CPU resources before utilizing all available I/O bandwidth. PlanetScale claims that its solution reduces latency and improves consistency, offering a reportedly more cost-effective alternative to equivalent setups on Amazon Aurora, Google Cloud SQL, or Supabase."
"Neki is our Postgres sharding solution. Built by the team behind Vitess combining the best of Vitess and Postgres. Neki is not a fork of Vitess. Vitess' achievements are enabled by leveraging MySQL's strengths and engineering around its weaknesses. Vitess is a clustering system developed at YouTube to scale MySQL horizontally from one to thousands of servers, each with its own local storage."
"While Vitess started as an open source project in 2010, the database infrastructure company has chosen a different path for Neki, and the source code is currently not available. To achieve Vitess' power for Postgres we are architecting from first principles and building alongside design partners at scale. When we are ready we will release Neki as an open source project suitable for running the most demanding Postgres workloads."
PlanetScale offers a managed sharded PostgreSQL service in general availability for AWS and Google Cloud. The service runs on PlanetScale Metal using local NVMe drives and claims "Unlimited I/O," with CPU rather than I/O typically becoming the bottleneck. PlanetScale asserts the setup reduces latency, improves consistency, and can be more cost-effective than Amazon Aurora, Google Cloud SQL, or Supabase equivalents. PlanetScale is developing Neki, a Postgres sharding solution from the Vitess team, which is not a fork of Vitess and whose source is not yet available but is planned for future open-source release.
Read at InfoQ
Unable to calculate read time
[
|
]