Google Cloud Launches Autopilot for Kubernetes Engine

Google Cloud has unveiled a new operational mode for its Kubernetes Engine (GKE), shifting the responsibility for numerous daily cluster management tasks to Google’s engineering teams and automated systems. Dubbed Autopilot, this new functionality handles all Day 2 operations related to cluster and node management, while simultaneously enforcing optimal practices for operation and security.

Augmenting the Existing GKE Experience

This new mode complements the current GKE functionality, which already manages the majority of the infrastructure required for cluster deployment. The previously existing experience, now referred to as the “standard” mode by Google Cloud, remains available, allowing users extensive customization options and manual control over their node infrastructure.

Drew Bradstock, Group Product Manager for GKE, explained that Autopilot was conceived to consolidate Google’s existing GKE tools with the expertise of its Site Reliability Engineering (SRE) teams, who have extensive experience running these clusters in a production environment within the company.

Key Features of Autopilot

“Autopilot integrates auto-scaling, automatic upgrades, maintenance procedures, and Day 2 operations, and crucially, it does so in a secure and robust manner,” Bradstock stated. “This allows our early adopters to rapidly provide an improved environment for developers, testing, and production workloads, as they can achieve Day 2 operational readiness immediately following the five-minute cluster creation process.”

Image Credits: Google

From a developer’s standpoint, the core workflow remains unchanged. However, this new mode empowers teams to concentrate on their applications rather than the intricacies of Kubernetes cluster administration. Businesses benefit from the advantages of Kubernetes without the burden of routine management and upkeep.

This shift represents a broader trend within the Kubernetes ecosystem, as few organizations view Kubernetes management as a primary source of competitive advantage.

Pricing and Service Level Agreements

Utilizing Autopilot incurs additional costs beyond the standard GKE flat fee of $0.10 per hour per cluster (a free GKE tier offering $74.40 in billing credits is also available), alongside charges for the resources consumed by your clusters and pods. Google guarantees a 99.95% Service Level Agreement (SLA) for the Autopilot control plane and a 99.9% SLA for Autopilot pods deployed across multiple zones.

Image Credits: Google

Integration with Google Cloud’s Container Portfolio

Autopilot for GKE expands Google Cloud’s suite of container-focused products, which also includes Anthos for multicloud deployments and Cloud Run, Google’s serverless compute platform. Bradstock clarified, “Autopilot is about unifying the automation capabilities within GKE for running on Google Cloud, presenting them in a user-friendly package.”

“This is particularly beneficial for those new to Kubernetes or managing extensive fleets, as it significantly reduces the time, operational overhead, and compute resources required.”

Distinction from Anthos and Cloud Run

While GKE is a fundamental component of Anthos, that service focuses on extending Google’s configuration management, service mesh, and other tools to an enterprise’s on-premises data centers. Currently, Autopilot for GKE is exclusively available on Google Cloud.

“Cloud Run excels in providing a streamlined development experience on the serverless side,” Bradstock added. “It enables rapid scaling from zero to 1,000 and back without requiring any management overhead. Autopilot, conversely, simplifies the entire platform for those who want to leverage the full Kubernetes ecosystem, maintain greater control, and manage numerous applications within a single environment.”