Four Powerful, New Features to Help You Build and Deploy More Efficient Apps On DigitalOcean Kubernetes

We’re adding to March’s updates with even more Managed Kubernetes features that will help you get even more utility out of the product–including newly supported Droplet types, the ability to automatically scale nodes to zero when you’re not using them, and more. Let’s walk through these new features and how they can benefit both your Kubernetes environment and your business.

DigitalOcean Managed Kubernetes support for new GPU Droplets types

You can now deploy GPU-accelerated workloads on DigitalOcean Managed Kubernetes using our latest GPU Droplet types (both NVIDIA and AMD). These new instance types are ideal for AI/ML training and inference, image and video processing, and other compute-intensive workloads. With native support for GPU nodes in your Kubernetes clusters, you get the flexibility of containers with the raw power of high-performance GPUs-fully integrated into the DOKS experience. Here are the new GPU Droplet types:

Nodepool Scale-to-zero

This feature allows a node pool to automatically scale down to zero nodes when there are no active workloads that require those nodes. You can now enable the node pools within your Kubernetes environment to automatically scale down to zero when idle, stopping compute charges during those periods of inactivity. This feature is optimal for development or testing environments, applications with usage patterns tied to business hours that naturally yield idle periods, or workloads that use specialized node pools like GPU or CPU-Optimized for intermittent jobs. The main components of this feature include:

1. Reduce node pools to zero

• Allows you to set the minimum node count (min-nodes) to 0 using the UI, CLI, or API.

• Seamless integration with your existing autoscaling configurations.

• Maintain full control over which node pools are able to scale to zero, giving you more customization and control.

2. Automatic scaling

• Automatically detects pending pods that require scaled-down resources, helping customers to efficiently allocate resources without impacting availability.

• Conversely, when workloads are scheduled that require the node pool, it automatically scales back up. The Cluster Autoscaler automatically detects pending pods that require a scaled-down node pool and provisions the necessary nodes on demand.

• Provisions nodes on-demand when workloads are scheduled, allowing for dynamic infrastructure adjustments.

3. Cost optimization

• Eliminates compute charges for idle node pools, saving you money.

• Valuable for development, testing, and specialized workloads where resource demands fluctuate.

• Enables a true pay-per-use infrastructure model within Kubernetes, aligning costs directly with consumption.

Now you can enable true on-demand infrastructure, and automatically ensure you only pay for compute resources when they’re actually in use, whenever you need them.

DigitalOcean Kubernetes now available in our new, AI-optimized Atlanta data center (ATL1)

DigitalOcean’s newest AI-optimized data center, ATL1 in Atlanta-Douglasville, is now fully operational, so you can now deploy fully-managed Kubernetes clusters in the southeast United States.

As our newest and largest data center, ATL1 it is purpose-built to deliver high-density GPU infrastructure optimized for AI/ML workloads. For latency-sensitive applications, AI/ML inference workloads, and regional deployments—this means faster response times, reduced data transfer delays, and better performance.

DigitalOcean Kubernetes Routing Agent now generally available (GA)

We’re excited to announce the general availability (GA) of the DOKS routing agent: a fully managed solution that simplifies static route configuration in your Kubernetes clusters. With support for Kubernetes custom resources, this tool makes it easy to define custom routes, use Equal-Cost Multi-Path (ECMP) routing across multiple gateways, and override default routes without disrupting connectivity. To learn more about this feature, check out our blog post from March. You can also target routes to specific nodes using label selectors, making this ideal for use cases like VPN integration, custom egress paths, and self-managed VPC gateways.

We’re excited to introduce these new features, designed to help you build, deploy, and scale your applications faster and more efficiently–while expanding what’s possible on Kubernetes.

Get started

• Spin up a Kubernetes cluster

• Explore our product documentation

• Check out our Managed Kubernetes homepage

• Log into your DigitalOcean account**