This page shows you how to deploy and scale workloads more quickly in
Google Kubernetes Engine (GKE) clusters using fast-starting nodes. Fast-starting
nodes are used in GKE with Autopilot mode on a
best-effort basis when workloads use compatible configurations.

Fast-starting GKE nodes have significantly lower startup time for
compatible machine families. The accelerated startup time provides you with the
following benefits:

- Faster cold start
- Faster autoscaling
- Improved Pod scheduling long-tail latency
- Improved infrastructure cost efficiency

With fast-starting nodes, GKE pre-initializes hardware resources
to accelerate startup time. The pre-initialized resources are available on a
best-effort basis. Surge requests might only be partially served. Without
fast-starting nodes, resources are initialized on-demand, and nodes are served
at normal startup time.

## Requirements

Fast-starting nodes require no additional configuration. GKE
automatically uses fast-starting nodes if your workloads use compatible
configurations. You must meet all of the following requirements to use
fast-starting nodes:

- Use Autopilot clusters, or run [workloads in Autopilot
  mode in your Standard
  clusters](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/about-autopilot-mode-standard-clusters).
- Don't use any features that are incompatible with fast-starting nodes. For more information, see [Limitations](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/fast-starting-nodes.md.txt#limitations).
- Use one of the following compute resources which are described in the
  respective tables in the [Configuration requirements for eligible compute resources](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/fast-starting-nodes.md.txt#requirements-eligible) section:

  - [Autopilot container-optimized compute platform](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/fast-starting-nodes.md.txt#autopilot-compute-platform-requirements)
  - [G2 machine series](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/fast-starting-nodes.md.txt#g2-requirements)
  - [A2 machine series](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/fast-starting-nodes.md.txt#a2-requirements)
  - [A3 machine series](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/fast-starting-nodes.md.txt#a3-requirements)
  - [G4 machine series](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/fast-starting-nodes.md.txt#g4-requirements)

## Limitations

The following features aren't compatible with fast-starting GKE
nodes. If you use any of these features, GKE provisions nodes
with the typical startup time:

- G2 with `DEFAULT` GPU driver version on GKE versions earlier than 1.33.0-gke.1304000. For 1.33.0-gke.1304000 or later, both `LATEST` and `DEFAULT` work with fast-starting nodes.
- [Secondary boot
  disks](https://docs.cloud.google.com/kubernetes-engine/docs/how-to/data-container-image-preloading) on versions earlier than 1.33.2-gke.1015000.
- [Spot VMs](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/spot-vms.md.txt)
- [Placement policies](https://docs.cloud.google.com/compute/docs/instances/placement-policies-overview.md.txt)
- [Multi-network
  support](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/about-multinetwork-support-for-pods)

## Autopilot GPU workloads

Requesting compatible GPUs in Autopilot mode results in up to four
times faster node startup time and up to two times faster Pod scheduling time
than similar requests in GKE Standard mode, because the
Autopilot GPU workloads can use fast-starting nodes.

The following are some example use cases. However, any Pods meeting the
conditions from the [Requirements](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/fast-starting-nodes.md.txt#requirements) section are compatible with
fast-starting nodes.

### ComputeClass

Request a compatible accelerator type and count in a ComputeClass, like in
the following example:

```yaml
apiVersion: cloud.google.com/v1
kind: ComputeClass
metadata:
  name: ACCELERATOR_COMPUTE_CLASS_NAME
spec:
  priorities:
  - gpu:
      type: ACCELERATOR_TYPE
      count: ACCELERATOR_COUNT
  nodePoolAutoCreation:
    enabled: true
```

When you select this ComputeClass in a Pod, like in the following example,
GKE uses fast-starting nodes:

```yaml
apiVersion: v1
kind: Pod
metadata:
  name: POD_NAME
spec:
  nodeSelector:
    # Select a ComputeClass that requests compatible GPUs
    cloud.google.com/compute-class: ACCELERATOR_COMPUTE_CLASS_NAME
  containers:
  - name: my-container
    image: registry.k8s.io/pause
    resources:
      limits:
        nvidia.com/gpu: ACCELERATOR_COUNT
```

Replace the following values:

- `ACCELERATOR_COMPUTE_CLASS_NAME`: the name of the ComputeClass that requests the accelerators.
- `ACCELERATOR_TYPE`: the type of accelerator.
- `ACCELERATOR_COUNT`: the number of accelerators required by the Pod. This value must be less than or equal to the value in the `spec.priorities.gpu.count` field in the ComputeClass.
- `POD_NAME`: the name of your Pod.

For more information about ComputeClass, see [About custom compute
classes](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/about-custom-compute-classes).

### Pod specification

Select a compatible accelerator type and count in your Pod specification, like
in the following example:

```yaml
apiVersion: v1
kind: Pod
metadata:
  name: POD_NAME
spec:
  nodeSelector:
    cloud.google.com/gke-accelerator: ACCELERATOR_NAME
  containers:
  - name: my-container
    image: registry.k8s.io/pause
    resources:
      limits:
        nvidia.com/gpu: ACCELERATOR_COUNT
```

Replace the following values:

- `POD_NAME`: the name of your Pod.
- `ACCELERATOR_NAME`: the name of the accelerator required by the Pod.
- `ACCELERATOR_COUNT`: the number of accelerators required by the Pod.

## Configuration requirements for eligible compute resources

The tables in the following sections describe the specific requirements for
GKE to provision the relevant compute resources as fast-starting
nodes.

### Autopilot container-optimized compute platform

See the following requirements for GKE to provision fast-starting
nodes with the Autopilot container-optimized compute platform.

| Compute resource | Release channel and version support | Supported disks |
|---|---|---|
| [Autopilot container-optimized compute platform](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/autopilot-overview.md.txt#autopilot-compute-platform) | - Rapid channel - 1.33.0-gke.1696000 or later | - `pd-balanced` boot disks up to 100 GiB - No [Local SSDs](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/local-ssd.md.txt). |

### G2 machine series

See the following requirements for GKE to provision G2 machines
as fast-starting nodes.

| Compute resource | Release channel and version support | Supported disks |
|---|---|---|
| [G2 machine series](https://docs.cloud.google.com/compute/docs/accelerator-optimized-machines.md.txt#g2-vms) | - Rapid and Regular channels - 1.31 or later | - `pd-balanced` boot disks up to 500 GiB - No [Local SSDs](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/local-ssd.md.txt). |

### A2 machine series

See the following requirements for GKE to provision A2 machines
as fast-starting nodes.

| Compute resource | Release channel and version support | Supported disks |
|---|---|---|
| [A2 Standard machine types](https://docs.cloud.google.com/compute/docs/accelerator-optimized-machines.md.txt#a2-standard-vms) | - Rapid and Regular channels - 1.32.0-gke.1008000 or later | - `pd-balanced` or `pd-ssd` boot disks up to 1,000 GiB. - No [Local SSDs](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/local-ssd.md.txt). |
| [A2 Ultra machine types](https://docs.cloud.google.com/compute/docs/accelerator-optimized-machines.md.txt#a2-ultra-vms) | - Rapid and Regular channels - 1.32.0-gke.1008000 or later | - `pd-balanced` or `pd-ssd` boot disks up to 1,000 GiB, with higher limits for `a2‑ultragpu‑4g` (1,500 GiB) and `a2‑ultragpu‑8g` (3,000 GiB). - With automatically attached [Local SSDs](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/local-ssd.md.txt). |

### A3 machine series

See the following requirements for GKE to provision A3 machines
as fast-starting nodes.

| Compute resource | Release channel and version support | Supported disks |
|---|---|---|
| [A3 Mega machine type](https://docs.cloud.google.com/compute/docs/accelerator-optimized-machines.md.txt#a3-mega-vms) | - Rapid channel - 1.34.4-gke.1130000 or later | - `hyperdisk-balanced`, `pd-ssd`, or `pd-balanced` boot disks. - With automatically attached [Local SSDs](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/local-ssd.md.txt). |
| [A3 High machine type](https://docs.cloud.google.com/compute/docs/accelerator-optimized-machines.md.txt#a3-high-vms) (`a3-highgpu-8g`) | - Rapid channel - 1.34.4-gke.1130000 or later | - `hyperdisk-balanced`, `pd-ssd`, or `pd-balanced` boot disks. - With automatically attached [Local SSDs](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/local-ssd.md.txt). |

### G4 machine series

See the following requirements for GKE to provision G4 machines
as fast-starting nodes.

| Compute resource | Release channel and version support | Supported disks |
|---|---|---|
| [G4 machine series](https://docs.cloud.google.com/compute/docs/accelerator-optimized-machines.md.txt#g4-series) | - Rapid channel - 1.34.4-gke.1130000 or later | - `hyperdisk-balanced` boot disks - No [Local SSDs](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/local-ssd.md.txt). |

## Pricing

Fast-starting nodes are available in GKE Autopilot at no
extra charge. For more information about GKE Autopilot
pricing, see the Autopilot mode section in [Google Kubernetes Engine
pricing](https://cloud.google.com//kubernetes-engine/pricing#autopilot_mode).

## What's next

- [Strategies to reduce startup latency](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/startup-latency-reduction-strategies.md.txt)
- [About the Autopilot container-optimized compute
  platform](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/autopilot-overview#autopilot-compute-platform)
- [About custom compute
  classes](https://docs.cloud.google.com/kubernetes-engine/docs/concepts/about-custom-compute-classes)
- [Persistent Disk performance overview](https://docs.cloud.google.com/compute/docs/disks/performance.md.txt)