We are happy to unveil the newest release of Crunchy Postgres for Kubernetes version 5.4. This update brings an array of features set to improve your experience including:
- Support for ARM
- Native vector search via pgvector
- Comprehensive support for huge pages
- Native support for Postgres tablespaces
- Documentation enhancements
To explore version 5.4, head over to our Developer Portal Quickstart to test it, or our Customer Portal to download the latest version for your production needs. To see Crunchy Postgres for Kubernetes in action, managing your production Postgres workloads, don't hesitate to request a demo. Let’s dig in more to what’s included within this release.
ARM support for improved price to performance ratio
As Postgres needs to support an array of operating systems, it leverages a platform-agnostic approach for everything from disk I/O to memory management. For many years, this strategy has enabled Postgres to take full advantage of the rapid advancements in Linux on x86 processors. Today, we at Crunchy Data are thrilled to announce that we can offer comparable performance on ARM-based systems using Crunchy Postgres for Kubernetes. These systems deliver a more cost-effective solution with superior power efficiency.
Let’s take a look at performance on two comparable EC2 instances:
- c7g.xlarge powered by AWS Graviton3E with 4 vCPU’s, 8GB of RAM and an average monthly cost of $105
- c6i.xlarge powered by 3rd Generation Intel Xeon Scalable processors with 4 vCPU’s, 8GB of RAM and an average monthly cost of $124
The Crunchy Postgres for Kubernetes pods were configured with four vCPU’s and 8GB of RAM. All of the underlying EC2 instances utilized similar disk and networking configurations.
Even though the ARM64 instances were roughly 18% less expensive, performance results across both instance types were nearly identical. Even more striking was the fact that the ARM databases slightly outperformed the x86 databases under higher connection counts on the larger dataset.
As you can see, on the larger dataset the ARM based database started to outperform the x86 based database at higher numbers of clients and transactions per second. This illustrates that ARM-based systems can deliver robust performance, with the added benefits of increased power efficiency and cost-effectiveness.
Huge pages and Postgres in containers
Huge pages are a Linux kernel feature to increase CPU efficiency in high-memory environments. They improve system performance and protect processes from the Out Of Memory (OOM) manager in the Linux kernel. Anyone who adjusts Postgres shared_memory should consider tuning their system’s huge pages.
Successfully configuring huge pages through all the Kubernetes and container layers is intricate! It's essential when tuning Postgres, so we've ensured that Crunchy Postgres for Kubernetes handles it for you. In addition, we're working with the broader community to refine the management of huge pages in Kubernetes and containers, aiming for a better experience for everyone.
Tablespace support added
With the launch of Version 5.4, we're proud to introduce tablespace support in Crunchy Postgres for Kubernetes. Why does it matter? Two words: flexibility and performance. You can organize your disk space more effectively, potentially speeding up your database. It is also about giving you more control over your data storage and management, making your life easier.
Vector support in Crunchy Postgres for Kubernetes
We have been talking with a lot of you who are out there building the future with your new applications and services. It is no surprise with the huge wave of interest in Language Learning Models (LLMs) and things like OpenAI's ChatGPT.
pgvector is a Postgres extension that lets you store, index, and search vectors efficiently. Think of vectors like an array of numbers representing a concept. They are very important in AI for representing and understanding complex connections and relationships to different concepts. It's like giving your AI a superpower to understand the nuances of language, much like we humans do!
We've taken pgvector and bundled it right into Crunchy Postgres for Kubernetes 5.4. Now you have the power to feed your Language Learning Models with Postgres, right there running on Kubernetes. You can store massive amounts of text data or vectors in your Postgres database, and use this data to train or operate your LLMs.
What’s new with docs?
In addition to all the feature improvements in Crunchy Postgres for Kubernetes 5.4, we have made some big changes to our documentation as well. First is a new theme which should allow for easier navigation, but we’ve also made some big improvements to content available to you there:
As part of all these changes our documentation will be moving to a separate repo. More to come on this in the future, but this will allow us to automatically deploy from main which means faster docs changes and a better experience for you.
We're incredibly excited to bring you version 5.4 of Crunchy Postgres for Kubernetes. These highlights offer a glimpse into the features and fixes included in this release. We're eager to hear your thoughts and we welcome your feedback. The full updated notes and detailed feature list are available in our documentation. Enjoy exploring the new capabilities!
July 11, 2023 •More by this author