Google introduced Kubernetes in 2014 as an open-source initiative to streamline container management. As Kubernetes has evolved into a crucial component of cloud-native technology, Google has consistently supported the project while also providing its commercial variant, Google Kubernetes Engine (GKE). Today, during Google Cloud Next, the company unveiled a new enterprise version of GKE—GKE Enterprise.
Chen Goldberg, General Manager and VP for cloud runtimes at Google, describes GKE Enterprise as the culmination of years of advancements in container management. In 2019, Google launched Anthos, a platform that facilitates seamless workload migration between different cloud environments. According to Goldberg, GKE Enterprise merges Anthos with GKE, streamlining the management of complex workloads for enterprises operating multiple clusters.
GKE Enterprise comes equipped with several advanced features tailored for intricate Kubernetes environments. These include security and governance tools, service mesh management, and a comprehensive dashboard for monitoring all workloads across the organization.
An exciting new concept introduced is the management of "fleets of clusters." This feature allows development teams to operate independently while adhering to shared corporate guidelines for cluster management. “Teams can implement policies, customize standard configurations for development, staging, and production environments, and monitor cost usage and vulnerabilities,” Goldberg explained. The management capabilities extend across multiple Kubernetes projects via a unified tool.
Furthermore, Google enables customers to organize their clusters hierarchically and set granular rules as needed. Goldberg elaborated, “As a GKE administrator or platform team, I can create fleets of clusters, manage them collectively, and establish a new concept called teams, granting specific permissions for those fleets.”
In tandem with these management enhancements, Google also announced a new chip designed for artificial intelligence workloads—the Cloud TPU v5e. "TPU v5e is unique because it can scale to tens of thousands of chips, making it perfect for developing intricate AI models," Goldberg stated. This chip will be available for preview in GKE.
“GKE stands out as the most scalable managed Kubernetes service available. It supports clusters of up to 15,000 nodes, complete with automatic updates and workload orchestration capabilities. Many features of Kubernetes and GKE align seamlessly with the latest developments in generative AI, leveraging TPUs and GPUs,” she added.
To further integrate generative AI, Google is deploying a Large Language Model (LLM) trained on its own documentation to assist GKE users. Through Duet AI for GKE and Cloud Run, users can ask questions and receive clear, straightforward answers based on existing documentation.
“It provides script examples and helps users to code more efficiently. Since the model is trained on our documentation and code samples, it enhances the relevance and quality of the responses,” Goldberg noted. However, it’s crucial to recognize that LLMs have limitations, including the risk of producing inaccurate answers when faced with insufficient information. Even a constrained dataset doesn’t completely mitigate this issue.
GKE Enterprise is set to enter preview in September, while Cloud TPU v5e will be available for preview this week. Duet AI for GKE and Cloud Run is included in the expanded Duet AI preview from Google Cloud.
Ranking
