Back to Top

Selecting a deployment pattern for imagery data management systems

Imagery data management systems are typically deployed using one of four deployment patterns:

Selecting a deployment pattern is one of the most important decisions to make in designing a GIS system for your organization.

The most important factor in this decision will be aligning with your organization’s IT principles, guidelines, and comfort-level in supporting different deployment approaches. For example, some organizations may prefer to standardize on SaaS-based systems and solutions. Other organizations that are investing heavily into Kubernetes-based deployments, including hiring and training staff with operational Kubernetes experience and skills, may prefer Kubernetes-based deployment patterns.

Note:

The capabilities as well as the considerations differ between deployment patterns. Review the comparisons below, along with the deployment pattern pages for additional information.

For general information and considerations around these deployment approaches see the ArcGIS products and deployment options page of the ArcGIS overview.

Capability comparison

In addition to aligning with your IT principles, guidelines, and comfort-level, it’s also important to consider the capabilities of each deployment pattern in your decision-making process. The capabilities of an imagery data management system differ between deployment patterns. The following matrix compares the specific capabilities supported by each of the deployment pattern:

Capabilities used in an imagery data management system, but typically provided by other systems, such as basemaps, geocoding, and other location services provided by a location services system are not listed below.

Capability                         SaaS SaaS - Image Dedicated   Windows/Linux Kubernetes
Imagery visualization and analysis
Data modeling and structuring
Imagery data publishing 1 1
Tiled imagery layer hosting2    
On-the-fly raster analysis
Elevation analysis
Distributed raster analytics
Image extraction
Deep learning and AI 3 3 3
Multidimensional data
Work in image space 4 4 4
Work with stereo imagery
Oriented imagery
Drone operations
Reality mapping
Video5      
Lidar datasets and workflows
Synthetic aperture radar
STAC catalogs6

Full support

Partial support

  1. ArcGIS Online systems can support a wide variety of imagery data formats but may not support as many types, sensors and products as ArcGIS Enterprise-based systems. ↩ ↩2

  2. Tiled imagery layers refer to a specific, CRF-based tile layer in ArcGIS Online, see Tiled imagery layers for more information. ↩

  3. ArcGIS Deep Learning Studio is a web application that is only available with ArcGIS Enterprise on Windows or Linux at this time. ↩ ↩2 ↩3

  4. ArcGIS Excalibur is a web application that is one of the primary applications for working with imagery in image space, as well as video analysis. It is only available for ArcGIS Enterprise at this time. Image space can still be visualized and exploited in ArcGIS Pro. ↩ ↩2 ↩3

  5. Cataloging and streaming of video files is currently only available with ArcGIS Video Server, a federated server role on Windows or Linux. ArcGIS Pro and ArcGIS Excalibur both support use of full motion video stored on network, cloud or web storage or streamed through Video Server. ↩

  6. Spatio-temporal Asset Catalog (STAC) support is available in ArcGIS Pro to connect to external STAC catalogs, no deployment pattern currently supports hosting STAC catalogs, though ArcGIS GeoPortal Server can be used to host a STAC catalog. ↩

See the imagery data management system capabilities for more information on each row listed above. Additionally, each of the cells above is described in more detail in the imagery data management system deployment pattern pages.

The capabilities represented above reflect those available as of December, 2024.

General considerations

The considerations below aim to help align your organization’s business and IT needs with the appropriate imagery data management system deployment pattern. The information presented here is not meant to be exhaustive, but rather highlights key considerations for designing and implementing this system pattern.

  • Scalability, reliability, service level agreements (SLA), security, and the balance of responsibility between your organization and Esri tend to be major factors in selecting a deployment pattern. See the reliability, performance and scalability, and security pillars for more information.
  • The SaaS deployment patterns using ArcGIS Online along with ArcGIS Image or ArcGIS Image Dedicated feature the quickest time to market. The capabilities they provide can be enabled quickly for an organization, and the variety of applications included support a wide range of workflows and user needs.
  • ArcGIS Image for ArcGIS Online is a good option for organizations working with relatively simpler or static imagery assets, such as a dataset delivered from a vendor or downloaded from a data provider. While there is some time-enabled support through tiled imagery layers, ArcGIS Image datasets are usually self-contained and do not require the complex query and mosaicking that other workflows may require.
  • ArcGIS Image Dedicated is designed for more advanced imagery data management systems, and one key differentiator from ArcGIS Image is that Dedicated supports the use of customer-managed object storage in AWS or Azure as a source for imagery data. Check with the product documentation and example workflows to verify compatibility with your storage provider, cloud hosting strategy and implementation details.
  • Imagery data management systems built with ArcGIS Enterprise on Windows, Linux, or Kubernetes support connections to object storage in a wide variety of cloud environments as well as on-premises object storage and block storage providers.
  • Access to a Graphics Processing Unit (GPU) is frequently an important part of imagery data management systems, both in terms of the client applications and the server or backend systems.
    • Deep learning workflows such as inferencing or training are heavily reliant on specific computing resources (e.g. a GPU processor). Esri provides a deep learning FAQ that answers common questions related to GPU availability and suitability for deep learning.
    • Some analytics workflows or rendering processes, such as viewing imagery with ArcGIS Pro on a virtual machine, rely on a GPU or virtual GPU to accelerate this experience, while other workflows such as geoprocessing tools or Python Notebooks rely on a GPU for native analytical calculations. See GPU processing with Spatial Analyst in the ArcGIS Pro documentation for additional details.
    • The availability and utilization of GPU resources can vary significantly between different deployment patterns, and can also have a significant impact on the cost of hosting a system. Specific guidance for ArcGIS Enterprise on Kubernetes is available in the relevant documentation.
    • ArcGIS Notebook Server and ArcGIS Online Notebooks provide a method to giving users access to GPU resources using specially-defined notebook runtimes, which can assist with data science and machine learning workflows in Python notebooks.
    • Note that access to a GPU is not a universal requirement – many geoprocessing tools or analytical workflows are not designed to, or able to take advantage of a GPU, so do not over-provision this resource, hoping it will help the performance of any workflow.

Selecting a deployment pattern is one of the most important decisions to make in designing a GIS system for your organization. However, it is not the only one. There are many additional factors to take into consideration when designing your system, including areas like security, reliability, and integration. As such, don’t consider the information provided here to be exhaustive. Review the architecture practices and pillars of the ArcGIS Well-Architected Framework, as well as product documentation, in detail as part of your design process.

Top