Timescale recently introduced Dynamic PostgreSQL, a new cloud-managed option to scale the database capacity within a predefined vCPU range. Promoted as "buy the base and rent the peak", the new option scales the capacity according to the load, trying to address the unpredictability and variability of serverless options.
Based on TimescaleDB, the open-source time series database that extends PostgreSQL, Dynamic PostgreSQL wants to offer a third alternative to provisioned and serverless databases. Mike Freedman, CTO and co-founder at Timescale, and Grant Godeke, senior product manager at Timescale, explain:
It is backed by dynamic compute, a Timescale innovation that instantaneously scales your available compute within a predefined min/max range according to your load. Instead of provisioning for your peak (and paying for it at all times), you can now pick a compute range: your database will operate at base capacity and instantly access up to its peak only when needed. Buy the base, rent the peak.
As the customer chooses a range, the dynamic maximum is capped at twice the base capacity. Timescale argues that databases are very different from Lambda functions and serverless databases today are inefficient for most production workloads as they focus on the extremes for scaling up and down, and they introduce higher and hard-to-understand pricing to account for the resources reserved to serve changing demands. Jeremy Daly, CEO and founder of Ampt, writes:
The difference here (I think) is that they are positioning this as "Buy the base, rent the peak." I’ve been saying this for a very long time, but this is the missing piece of serverless that cloud providers need to embrace.
Tobias Petry, database consultant, comments:
It’s like burstable EC2 machines, which is a perfect solution: Cost-effective base pricing and you only pay more in the rare cases when needed. This effectively works against the oversized instances by default of teams.
One of the benefits of serverless databases is the ability to scale the capacity to zero, paying only for the compute time used. Freedman and Godeke argue:
There are some use cases where "scale to zero" makes sense. For example, proof-of-concept demos or more hobbyist applications (...) But for your production database and in more operational settings? You don’t want to scale to zero. Scaling to zero means a "cold boot" on restart: empty database shared buffers, empty OS cache, empty catalog caches.
Dynamic PostgreSQL targets mainly deployments running on AWS, claiming that customers will save 10-20 % when migrating from RDS for PostgreSQL and 50-70% when migrating from Aurora Serverless but no benchmarks have been published yet.
Dynamic PostgreSQL was not the only announcement during the Dynamic Infra Launch Week: the Terraform provider is generally available, hypertables are supported on Cloudflare Hyperdrive and Timescale cloud is now available in 7 AWS regions in Australia, Europe, the US, and Asia.
The monthly pricing starts at $87.60 USD for a 1-2 CPU range, with $0.12 USD changed hourly per CPU for usage above the committed base. Timescale offers a 30-day free trial for new accounts.