Google retools Kubernetes for AI agents instead of building a new stack
Agentic workloads are pushing Kubernetes into territory its creators never planned for. Agents generate untrusted code on the fly, sit idle for long stretches and churn through pods at a pace that strains the orchestrator’s control plane. Cloud providers are now rethinking how general-purpose a cluster should be. Google LLC’s answer is to stop treating […] The post Google retools Kubernetes for…
Kubernetes, originally designed for microservices, is facing new challenges as agentic workloads strain the orchestrator's control plane, according to Jago Macleod, director of engineering for Kubernetes and Google Kubernetes Engine at Google Cloud. These workloads generate untrusted code on the fly, sit idle for long periods, and consume resources rapidly.
To address these issues, Google is rethinking how to optimize each cluster for its specific workload, preventing the fragmentation of the open-source ecosystem. Google's solution is Agent Substrate, which runs AI agents in sandboxes directly on Kubernetes, donating the project to the Cloud Native Computing Foundation. This approach tackles the inefficiency of allocating sandboxes for idle agents, wasting CPU cycles during agent inactivity.
By customizing Kubernetes for agentic workloads, Google aims to maintain its compute fabric while ensuring agents do not disrupt the system. As agentic workloads grow, human judgment becomes increasingly valuable in managing these specialized tasks.
Written by urgent.news from SiliconANGLE's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
