ArcGIS Pro can be used in Amazon Web Services (AWS) cloud environments to provide different deployment options. The guidance in this topic helps ArcGIS Pro users and GIS administrators choose the most appropriate AWS service for their workflows..
Choose the Amazon Web Services platform
The information below describes the different Amazon Web Services platforms, the ideal scenario for each, and key benefits and limitations.
| Platform | Scenario | ArcGIS Pro benefit | Limitation |
|---|---|---|---|
| Amazon WorkSpaces Graphics.g4dn bundle | ArcGIS Pro users who need a fast, managed GPU desktop | Dedicated NVIDIA pass-through GPU, predictable performance | This platform has fixed stock-keeping units (SKUs) and less image customization. |
| Amazon AppStream 2.0 | Multiple ArcGIS Pro users streaming sessions from large group GPU-backed virtual machines | Scalable, browser-based access, GPU sharing for cost savings | This platform requires profile tuning and has limited offline use. |
Amazon EC2 GPU Instances (G4/G5) | ArcGIS Pro power users or advanced teams who need the same flexibility as a physical workstation | Full administrative rights, scalable CPU and GPU, and reusable templates for machine configurations | This platform requires maintenance for drivers, system updates, and scaling. |
System considerations for deploying ArcGIS Pro in Amazon WorkSpaces include the following:
- GPU memory: At least 4 GB GPU RAM is required. G5 (A10G) instances provide larger GPU buffers for complex 3D and raster workflows.
- RAM: 32 GB minimum, 64 GB or more recommended for lidar or raster-heavy analysis.
- Latency: Less than 200 ms round-trip time (RTT). The performance is more sensitive to latency than bandwidth.
- Throughput: 1.5 Mbps for light workloads, 3 Mbps for medium workloads, and 5 or more Mbps for process-intensive 3D workflows.
- Storage: It is recommended that Amazon FSx (Windows or NetApp ONTAP) be used for shared file geodatabases and FSLogix profiles. For large imagery, raster archives, and cloud data stores, Amazon S3 is recommended.
Deploy the Amazon WorkSpaces Graphics.g4dn bundle
The Amazon WorkSpaces Graphics.g4dn bundle provides Windows 11 desktops with dedicated NVIDIA GPUs. This is recommended for users who need a ready-to-use ArcGIS Pro environment without a lot of maintenance.
This platform supports workflows such as viewing and querying 2D maps, editing, moderate geoprocessing use, and small 3D scenes. It is not recommended for very large 3D city models, lidar classification, or deep learning workflows.
To deploy ArcGIS Pro on Amazon WorkSpaces, complete the following steps:
- In the AWS console, click Create a WorkSpace > Graphics.g4dn Bundle with Windows 11.
- Assign a user profile and start the instance.
- Install ArcGIS Pro 3.x and license the software.
- Run the ArcGIS Pro Performance Assessment Tool (PAT) to validate rendering and frame rate.
- Connect through the Amazon WorkSpaces client for everyday use.
Deploy Amazon AppStream 2.0
Amazon AppStream 2.0 delivers ArcGIS Pro as a streamed application instead of a full Windows desktop. Users start ArcGIS Pro directly in a browser or client with GPU acceleration provided in the backend. This option is ideal for larger organizations, universities, or seasonal projects with many ArcGIS Pro users.
This platform supports map reviewers, simple 2D projects, editing in 2D and 3D, and moderate raster analysis. It is possible to support heavy software use, though it is limited to 1 to 2 users per GPU for complex 3D workflows.
Below are recommendations for ArcGIS Pro:
- GPU-supported fleets: Graphics.g4dn (NVIDIA T4) or Graphics.g5 (NVIDIA A10G)
- Density guidance: Up to six users per GPU for light use, up to 3-4 users per GPU for medium use, and 1-2 users per GPU for heavy use.
- Profiles: Store ArcGIS Pro user settings on FSx for Windows or Amazon S3.
- Scaling: Configure the group of machines to automatically expand and shrink with user demand.
To deploy ArcGIS Pro on Amazon AppStream 2.0, complete the following steps:
- In the AWS console, create an Amazon AppStream 2.0 Fleet, and choose Graphics.g4dn or Graphics.g5.
- Build an image by installing ArcGIS Pro 3.x, GPU drivers, and the necessary extensions.
- Configure a user profile with persistence using FSx or S3.
- Publish ArcGIS Pro as an application in an AppStream stack.
- Assign stack access to users and groups.
- Users connect using a browser or AppStream client to directly stream ArcGIS Pro.
- Validate session density and performance with the ArcGIS Pro Performance Assessment Tool (PAT).
Deploy Amazon EC2 instances (G4/G5)
For ArcGIS Pro workflows, Amazon EC2 GPU instances provide Windows desktops with administrative rights. This model is suited for power users running complex visualization, lidar point clouds, or deep learning with ArcGIS Pro extensions.
This platform supports small 2D projects, cartography, 2D and 3D editing, raster analysis, and terrain visualization workflows. It also supports large 3D city scenes, lidar editing, and GPU-based deep learning.
Learn more about sizing requirements for Amazon EC2 instances
To deploy ArcGIS Pro on an Amazon EC2 instance, complete the following steps:
- Start an Amazon EC2 instance and select the GPU-supported G4dn or G5 type.
- Add an SSD-based Amazon Elastic Block Store (EBS) volume for geodatabases and cache the volume.
- Install NVIDIA GRID drivers from AWS Marketplace.
- Install ArcGIS Pro 3.x and license the software.
- Validate rendering and performance with the ArcGIS Pro Performance Assessment Tool (PAT).
- Save as an Amazon Machine Image (AMI) that can be used as a template.