|
| 1 | +--- |
| 2 | +title: About Docker Offload |
| 3 | +linktitle: About |
| 4 | +weight: 15 |
| 5 | +description: Learn about Docker Offload, its features, and how it works. |
| 6 | +keywords: cloud, build, remote builder |
| 7 | +--- |
| 8 | + |
| 9 | +Docker Offload is a fully managed service for building and running containers in |
| 10 | +the cloud using the Docker tools you already know, including Docker Desktop, the |
| 11 | +Docker CLI, and Docker Compose. It extends your local development workflow into a |
| 12 | +scalable, cloud-powered environment, so you can offload compute-heavy tasks, |
| 13 | +accelerate builds, and securely manage container workloads across the software |
| 14 | +lifecycle. |
| 15 | + |
| 16 | +Docker Offload also supports GPU-accelerated instances, allowing you to |
| 17 | +containerize and run compute-intensive workloads such as Docker Model Runner and |
| 18 | +other machine learning or data processing tasks that benefit from GPU. |
| 19 | + |
| 20 | +## Key features |
| 21 | + |
| 22 | +Docker Offload includes the following capabilities to support modern container |
| 23 | +workflows: |
| 24 | + |
| 25 | +- Cloud-based builds: Execute builds on remote, fully managed BuildKit instances |
| 26 | + with native support for multi-platform output. |
| 27 | +- GPU acceleration: Use NVIDIA L4 GPU-backed environments for machine learning, |
| 28 | + media processing, and other compute-intensive workloads. |
| 29 | +- Ephemeral cloud runners: Automatically provision and tear down cloud |
| 30 | + environments for each container session. |
| 31 | +- Shared build cache: Speed up build times across machines and teammates with a |
| 32 | + smart, shared cache layer. |
| 33 | +- Hybrid workflows: Seamlessly transition between local and remote execution |
| 34 | + using Docker Desktop or CLI. |
| 35 | +- Secure communication: Use encrypted tunnels between Docker Desktop and cloud |
| 36 | + environments with support for secure secrets and image pulling. |
| 37 | +- Port forwarding and bind mounts: Retain a local development experience even |
| 38 | + when running containers in the cloud. |
| 39 | +- VDI-friendly: Use Docker Offload in virtual desktop environments or systems that |
| 40 | + don't support nested virtualization. |
| 41 | + |
| 42 | +## Why use Docker Offload? |
| 43 | + |
| 44 | +Docker Offload is designed to support modern development teams working across |
| 45 | +local and cloud environments. It helps you: |
| 46 | + |
| 47 | +- Offload heavy builds and runs to fast, scalable infrastructure |
| 48 | +- Accelerate feedback loops in development and testing |
| 49 | +- Run containers that require more resources than your local setup can provide |
| 50 | +- Build and run AI apps with instant access to GPU-powered environments |
| 51 | +- Use Docker Compose to manage complex, multi-service apps that need cloud |
| 52 | + resources |
| 53 | +- Maintain consistent environments without managing custom infrastructure |
| 54 | +- Develop efficiently in restricted or low-powered environments like VDIs |
| 55 | + |
| 56 | +Docker Offload is ideal for high-velocity development workflows |
| 57 | +that need the flexibility of the cloud without sacrificing the simplicity of |
| 58 | +local tools. |
| 59 | + |
| 60 | +## How Docker Offload works |
| 61 | + |
| 62 | +Docker Offload replaces the need to build or run containers locally by connecting |
| 63 | +Docker Desktop to secure, dedicated cloud resources. |
| 64 | + |
| 65 | +### Building with Docker Offload |
| 66 | + |
| 67 | +When you use Docker Offload for builds, the `docker buildx build` command sends |
| 68 | +the build request to a remote BuildKit instance in the cloud, instead of |
| 69 | +executing it locally. Your workflow stays the same, only the execution |
| 70 | +environment changes. |
| 71 | + |
| 72 | +The build runs on infrastructure provisioned and managed by Docker: |
| 73 | + |
| 74 | +- Each cloud builder is an isolated Amazon EC2 instance with its own EBS volume |
| 75 | +- Remote builders use a shared cache to speed up builds across machines and |
| 76 | + teammates |
| 77 | +- Builds support native multi-platform output (for example, `linux/amd64`, |
| 78 | + `linux/arm64`) |
| 79 | +- Build results are encrypted in transit and sent to your specified destination |
| 80 | + (such as a registry or local image store) |
| 81 | + |
| 82 | +Docker Offload manages the lifecycle of builders automatically. There's no need to |
| 83 | +provision or maintain infrastructure. |
| 84 | + |
| 85 | +> [!NOTE] |
| 86 | +> |
| 87 | +> Docker Offload builders are currently hosted in the United States East region. Users in |
| 88 | +> other regions may experience increased latency. |
| 89 | +
|
| 90 | +### Running containers with Docker Offload |
| 91 | + |
| 92 | +When you use Docker Offload to run containers, a Docker Desktop creates a secure |
| 93 | +SSH tunnel to a Docker daemon running in the cloud. Your containers are started |
| 94 | +and managed entirely in that remote environment. |
| 95 | + |
| 96 | +Here's what happens: |
| 97 | + |
| 98 | +1. Docker Desktop connects to the cloud and triggers container creation. |
| 99 | +2. Docker Offload pulls the required images and starts containers in the cloud. |
| 100 | +3. The connection stays open while the containers run. |
| 101 | +4. When the containers stop running, the environment shuts down and is cleaned |
| 102 | + up automatically. |
| 103 | + |
| 104 | +This setup avoids the overhead of running containers locally and enables fast, |
| 105 | +reliable containers even on low-powered machines, including machines that do not |
| 106 | +support nested virtualization. This makes Docker Offload ideal for developers |
| 107 | +using environments such as virtual desktops, cloud-hosted development machines, |
| 108 | +or older hardware. |
| 109 | + |
| 110 | +Docker Offload also supports GPU-accelerated workloads. Containers that require |
| 111 | +GPU access can run on cloud instances provisioned with NVIDIA L4 GPUs for |
| 112 | +efficient AI inferencing, media processing, and general-purpose GPU |
| 113 | +acceleration. This enables compute-heavy workflows such as model evaluation, |
| 114 | +image processing, and hardware-accelerated CI tests to run seamlessly in the |
| 115 | +cloud. |
| 116 | + |
| 117 | +Despite running remotely, features like bind mounts and port forwarding continue |
| 118 | +to work seamlessly, providing a local-like experience from within Docker Desktop |
| 119 | +and the CLI. |
| 120 | + |
| 121 | +Docker Offload provisions an ephemeral cloud environment for each session. The |
| 122 | +environment remains active while you are interacting with Docker Desktop or |
| 123 | +actively using containers. If no activity is detected for about 30 minutes, the |
| 124 | +session shuts down automatically. This includes any containers, images, or |
| 125 | +volumes in that environment, which are deleted when the session ends. |
| 126 | + |
| 127 | +## What's next |
| 128 | + |
| 129 | +Get hands-on with Docker Offload by following the [Docker Offload quickstart](/offload/quickstart/). |
0 commit comments