--- title: GKE Pod snapshots | daily.dev description: Google Kubernetes Engine now offers Pod snapshots, a feature that saves the full running state of a workload — including CPU and GPU memory — and restores it... canonical: https://daily.dev/posts/gke-pod-snapshots-by32x9jvc twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: GKE Pod snapshots | daily.dev og:description: Google Kubernetes Engine now offers Pod snapshots, a feature that saves the full running state of a workload — including CPU and GPU memory — and restores it... og:url: https://daily.dev/posts/gke-pod-snapshots-by32x9jvc og:image: https://api.daily.dev/og/posts/By32x9jvc.png og:image:alt: GKE Pod snapshots og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # GKE Pod snapshots **[Google Cloud](https://daily.dev/sources/gcp)** · 5 min read · 0 upvotes · 0 comments ## Summary Google Kubernetes Engine now offers Pod snapshots, a feature that saves the full running state of a workload — including CPU and GPU memory — and restores it on demand. This targets cold-start problems in AI inference and agentic sandbox workloads, where loading large model weights or initializing isolated environments can take minutes. Benchmarks show up to 89% reduction in startup latency, with a 70B parameter model loading in 37 seconds and an 8B model in 15 seconds. Adoption is configured via declarative Pod snapshot CRDs, letting teams choose which Pods to snapshot, where to store data, and how to manage retention and restore behavior. Codeway's Retake platform reported cutting startup latency from a custom caching solution's 1 minute down to 8 seconds after adopting the feature, while also simplifying its codebase and reducing idle GPU costs. Though built for AI inference and sandboxes, the feature is workload-agnostic and can help any application with long initialization times, such as Java monoliths or game servers. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What is GKE Pod snapshots and how does it reduce AI model loading time? GKE Pod snapshots is a Google Kubernetes Engine feature that saves the full running state of a workload, including CPU and GPU memory, and restores it on demand for new replicas. Instead of every new replica downloading and loading model weights independently, it restores directly from a saved snapshot, cutting startup latency by as much as 89%. Benchmarks show a 70B parameter model loading in 37 seconds and an 8B parameter model in 15 seconds. _daily.dev surfaces updates like this for engineers cutting AI inference cold-start times on Kubernetes._ ### How do you configure which Pods get snapshotted in GKE Pod snapshots? Configuration is done through a declarative policy defined with Pod snapshot custom resource definitions (CRDs). The policy specifies which Pods to snapshot, where to store the data, snapshot retention for cost control, and whether to restore from the last snapshot automatically or specify an explicit snapshot during a new Pod deployment. Snapshots can be triggered at workload startup or on-demand during the Pod lifecycle. _Teams evaluating Kubernetes snapshot policies can track configuration details like this on daily.dev._ ### What startup latency improvement did Codeway's Retake platform see after adopting GKE Pod snapshots? Codeway's Retake, an AI photo editing platform, cut startup latency to 8 seconds using GKE Pod snapshots, down from the 1 minute achieved by a custom caching layer they had built for compiled artifacts on A3 H100 GPUs. Replacing that custom system also reduced maintenance overhead, simplified their codebase, and let them dynamically spin GPUs up and down to cut idle costs. _daily.dev helps engineers weighing custom caching against managed snapshot tooling compare real adoption results._ ## Similar posts on daily.dev - [GKE CPU startup boost: Faster pod starts, lower costs](https://daily.dev/posts/gke-cpu-startup-boost-faster-pod-starts-lower-costs-41twx2pz2) · Google Cloud · 0 upvotes · 0 comments - [NVIDIA Dynamo Snapshot: Fast Startup for Inference Workloads on Kubernetes](https://daily.dev/posts/nvidia-dynamo-snapshot-fast-startup-for-inference-workloads-on-kubernetes-eo8wvm8l4) · NVIDIA Developer · 1 upvotes · 0 comments - [The Hackathon Fix That Cut Our Storage Costs by 93%](https://daily.dev/posts/the-hackathon-fix-that-cut-our-storage-costs-by-93--fwfdp2gyw) · Cast AI · 0 upvotes · 0 comments - [Cut GPU inference cold start from 8 minutes to less than a minute](https://daily.dev/posts/cut-gpu-inference-cold-start-from-8-minutes-to-less-than-a-minute-vhoj8trdz) · The New Stack · 0 upvotes · 0 comments --- Tags: [#kubernetes](https://daily.dev/tags/kubernetes), [#gpu](https://daily.dev/tags/gpu), [#ai-inference](https://daily.dev/tags/ai-inference), [#gke](https://daily.dev/tags/gke) [View this post on daily.dev](https://daily.dev/posts/gke-pod-snapshots-by32x9jvc) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"GKE Pod snapshots","url":"https://daily.dev/posts/gke-pod-snapshots-by32x9jvc","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/gke-pod-snapshots-by32x9jvc"},"datePublished":"2026-09-21T16:14:33.343Z","dateModified":"2026-09-27T07:04:36.552Z","description":"Google Kubernetes Engine now offers Pod snapshots, a feature that saves the full running state of a workload — including CPU and GPU memory — and restores it...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/71a924f2b121958fb9f73a189cb1663d?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/71a924f2b121958fb9f73a189cb1663d?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Google Cloud","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Google Cloud","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/gcp","url":"https://daily.dev/sources/gcp"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/gke-pod-snapshots-by32x9jvc","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"kubernetes,gpu,ai-inference,gke","timeRequired":"PT5M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Google Cloud","item":"https://daily.dev/sources/gcp"},{"@type":"ListItem","position":3,"name":"GKE Pod snapshots"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/gke-pod-snapshots-by32x9jvc#faq","mainEntity":[{"@type":"Question","name":"What is GKE Pod snapshots and how does it reduce AI model loading time?","acceptedAnswer":{"@type":"Answer","text":"GKE Pod snapshots is a Google Kubernetes Engine feature that saves the full running state of a workload, including CPU and GPU memory, and restores it on demand for new replicas. Instead of every new replica downloading and loading model weights independently, it restores directly from a saved snapshot, cutting startup latency by as much as 89%. Benchmarks show a 70B parameter model loading in 37 seconds and an 8B parameter model in 15 seconds. daily.dev surfaces updates like this for engineers cutting AI inference cold-start times on Kubernetes."}},{"@type":"Question","name":"How do you configure which Pods get snapshotted in GKE Pod snapshots?","acceptedAnswer":{"@type":"Answer","text":"Configuration is done through a declarative policy defined with Pod snapshot custom resource definitions (CRDs). The policy specifies which Pods to snapshot, where to store the data, snapshot retention for cost control, and whether to restore from the last snapshot automatically or specify an explicit snapshot during a new Pod deployment. Snapshots can be triggered at workload startup or on-demand during the Pod lifecycle. Teams evaluating Kubernetes snapshot policies can track configuration details like this on daily.dev."}},{"@type":"Question","name":"What startup latency improvement did Codeway's Retake platform see after adopting GKE Pod snapshots?","acceptedAnswer":{"@type":"Answer","text":"Codeway's Retake, an AI photo editing platform, cut startup latency to 8 seconds using GKE Pod snapshots, down from the 1 minute achieved by a custom caching layer they had built for compiled artifacts on A3 H100 GPUs. Replacing that custom system also reduced maintenance overhead, simplified their codebase, and let them dynamically spin GPUs up and down to cut idle costs. daily.dev helps engineers weighing custom caching against managed snapshot tooling compare real adoption results."}}]} ```