GCP – GKE delivers breakthrough Horizontal Pod Autoscaler performance
At Google Cloud, we are committed to providing the fastest and most reliable Kubernetes platform, Google Kubernetes Engine (GKE). Today, we are excited to announce an improved Horizontal Pod Autoscaler (HPA), the Kubernetes feature that automatically updates workload resources to match demand. We rearchitected the HPA stack, resulting in a significant improvement in scaling performance. You can apply it to your environment with the new Performance HPA profile, which delivers:
- 2x faster scaling: Workloads now scale up more than twice as quickly, enabling faster response times and improved application performance.
- Improved metrics resolution: A new fast metrics path with improved metrics resolution allows for more granular scaling and reaction.
- Linear scaling to up to 1000 HPA objects: HPA now supports high-scale deployments with consistent performance, helping you run large-scale applications with confidence.
- aside_block
- <ListValue: [StructValue([(‘title’, ‘$300 in free credit to try Google Cloud containers and Kubernetes’), (‘body’, <wagtail.rich_text.RichText object at 0x3e3c8334e430>), (‘btn_text’, ‘Start building for free’), (‘href’, ‘http://console.cloud.google.com/freetrial?redirectpath=/marketplace/product/google/container.googleapis.com’), (‘image’, None)])]>
Why this matters
Faster HPA response times have been a common request from many GKE users, who frequently overprovision resources to account for delays in the autoscaling stack, sometimes resulting in higher costs and lower efficiency.
The Performance HPA profile addresses these challenges by:
- Minimizing waste: The new HPA profile enables more precise scaling, reducing the need for over-provisioning and optimizing resource utilization.
- Improving application responsiveness: Faster scaling helps ensure applications can quickly adapt to changing demands, delivering a better user experience.
- Increasing operational efficiency: The new profile streamlines operations by simplifying scaling management and reducing the need for manual intervention.
Many GKE customers welcome the change.
“With GKE’s Performance HPA profile, we’ve witnessed a remarkable boost in horizontal auto-scaling speed. In our tests with over 1000 HPA objects, workloads scaled up twice as fast. We’re excited to leverage this performance enhancement in our production environments.” – Sophy Cao, Senior Engineer, Spotify
Get started today
The Performance HPA profile is available now as a preview opt-in feature for both GKE Standard and GKE Autopilot. We encourage all GKE users to try the new Performance HPA profile and experience its significant performance improvements firsthand. It only takes a single gcloud command to opt-in — see the guide to enable this feature in your cluster today!
Read More for the details.
