AWS Public Sector Blog

Enhance operational agility and decision advantage with AWS Snowball Edge

For Defense agencies, maintaining a battlefield advantage requires getting inside the adversary’s decision cycle and staying informed of their activities and movements to rapidly detect, deter, and engage threats. In a data-dependent world, success belongs to the side with decision advantage: the ability to acquire data and make sense of a complex and adaptive environment, and act smarter and faster than the competition. Tools for swift and efficient decision-making are the keys to success because they address modern disparities on the battlefield.

The reasons for these disparities are well-known: legacy systems and applications are scattered across on-premises and multi-cloud locations that are coupled with geographically dispersed data sources, creating a network connectivity challenge. Compounding the issue is the traditional big data challenge of data volume generated daily from sensors, satellites, connected devices, and platforms that need to be processed in environments with limited compute power, network connectivity, and high latency.

Understanding global environments requires more than just more data – it requires live two- and three-dimensional maps, new support tools, improved processes, seamless connectivity, and better collaboration that can scale to the needs of the environment. Amazon Web Services (AWS) offers tools that help deliver these resources to the most remote edge locations. The AWS Snow Family is a series of edge devices that bring cloud computing and cloud storage to places with limited compute and network connectivity.

This blog post shows how to address these challenges of big data and accelerate time to data insights with machine learning with AWS Snowball Edge device deployment at the edge.

Watch now: An introduction to AWS Snowball Edge.

Overview of Snowball Edge deployment solution

The following solution is a deployment framework for putting military defense application capabilities in a box, which supports a wide range of applications for rapid deployment of complex computing power at the edge. This Snowball Edge deployment pattern can expand to many other use cases that would benefit greatly from the Snowball Edge device usage, such as:

  1. Leveraging geospatial data in disaster response scenarios – from hurricanes to forest fires. Access to images from before and after the disaster help response teams determine the extent of damages and create a real-time response plan. Snowball Edge can deploy quickly with local data to an area.
  2. Leveraging Snowball Edge deployment to locations that have locality issues – some nations may have classified imageries that would otherwise be illegal to deploy outside those countries in the AWS Cloud.

Figure 1. The solution’s deployed components, described below.

Figure 1. The solution’s deployed components, described in the following section.

Snowball Edge solution’s deployed components

1. AWS Snowball Edge is an 80TB data transfer device with on-board storage and compute capabilities. You can use Snowball Edge to move large amounts of data into and out of AWS, as a temporary storage tier for large local datasets, or to support local workloads in remote or offline locations. A Snowball Edge Compute Optimized with GPU device is identical to a Snowball Compute Optimized device, except for an installed NVIDIA Tesla V100 GPU, equivalent to the one available in the Amazon Elastic Compute Cloud (Amazon EC2) P3. Depending on the workloads and the number of users, customers can scale up with multiple Snowball Edge devices. For this solution example, the following workloads, provided by custom applications, are running in the device:

  • High resolution satellite imagery
  • Object detection/identification machine learning workload
  • Application to mesh 3D map data with the real time imagery to put situations under context
  • Application user interfaces to interact with the users

2. The Mobile SATCOM Antenna provides timely commercial imagery into the Snowball Edge device.

3. Additional datasets such as maps and historical imageries can be stored in Amazon EC2 or Amazon Simple Storage Service (Amazon S3) compatible endpoints to enhance the context awareness of the users.

4. Web apps and/or REST APIs serve the context aware data (e.g., 2D/3D images, targets) to the customers.

Snowball Edge deployment system workflows

Figure 2. System workflows for the Snowball Edge deployment solution, explained in more detail in the following solution walkthrough.

Figure 2. System workflows for the Snowball Edge deployment solution, described in the following section.

Solution walkthrough

  1. Analysts plan an imagery request through the Mobile SATCOM Antenna. The satellites are then tasked to capture raw imagery.
  2. The imagery is downlinked from the satellite to the Mobile SATCOM Antenna. The analyst then sends the imagery to the Snowball Edge device for further processing.
  3. Image enhancement, such as object detection, cloud removal, and color corrections, are applied to the data through the machine learning (ML) software deployed on to the AWS Snowball device. One deployment option is AWS IoT Greengrass with an AWS Snowball device. AWS IoT Greengrass is an open-source Internet of things (IoT) edge runtime and cloud service that helps you build, deploy, and manage IoT applications on your devices. Additional data such as enhanced images are generated by ML software deployed on the device.
  4. Auxiliary datasets are used by the image processing in Step 3. They are also presented alongside the downlinked imagery to improve accuracies of the information presented to the analysts.
  5. Analysts visualize data in both 2D and 3D maps with annotations and context-aware information to aid in decision making.

Benefits of the Snowball Edge solution

1. Faster decisions with more concise datasets

Delayed or disconnected, intermittently-connected, and low-bandwidth (DIL) environments are a challenge for many on-premises or even cloud deployed applications. This solution addresses the concern by switching from big data to fast data. Because this Snowball Edge device solution is deployed to a specific location to monitor for a focused area or localized region, the traditional challenge related to data volume is minimized. Context-related data (e.g., auxiliary data) are pre-loaded to a Snowball Edge device as part of the machine image; this allows the system to process only a small amount of real-time data from external data sources to provide immediate insights.

2. Accelerate access to data and analysis

From sensors to data processing, all the systems mentioned above are very close in proximity. This allows images to be processed much faster with very little network latency. Traditionally, it takes hours to move downlink imagery through the processing pipelines and become accessible to users. With this deployment, because of fast data and reduced latency, processed real-time data is available for consumption in less than 15 minutes, which means insights are provided to the users ten or fifteen times faster than through traditional means. Shortening the time from data capture to processing enables users to make more accurate real-time decisions on the spot which brings compute closer to the data

3. Customizable to fit specific deployment capability requirement

Speed is essential and not only applies to the inference time for the object detection process. In a rapid deployment scenario, system deployment is critical to support operations. With a Snowball Edge device deployment, users can customize their deployment scenarios by picking from a list of custom pre-built Amazon Machine Images (AMIs) with all the necessary tools already built in for different workloads. Within the AWS Snowball console, customers can place an order for an AWS Snowball Edge device with the selection of just a few buttons. This ready-to-use deployment can be put into operations within days compared to the traditionally large system deployment that normally takes months.

Figure 3. The Snowball Edge order and delivery timeline, which helps meet an agile deployment schedule.

Figure 3. The Snowball Edge order and delivery timeline, which helps meet an agile deployment schedule.


To improve mission agility, you can now package mission-ready AMIs containing solution specific deployment. These AMIs can be deployed on-demand to the AWS Snowball to specific locations to support complex and adaptive environments. With the introduction of Amazon Elastic Kubernetes Service (Amazon EKS) Distro support at the edge, customers can run containerized applications on the AWS Snowball devices. This makes it even more simple for customers to standardize operations across all their environments and provide better consistency, flexibility, and portability to maintain battlefield advantage.

For next steps, take a look at AWS Snow FamilyAWS Snowball resources, and the AWS Snowball FAQs. For full details around configuration options and pricing, see AWS Snowball Pricing.

Read more about AWS for the edge:

Subscribe to the AWS Public Sector Blog newsletter to get the latest in AWS tools, solutions, and innovations from the public sector delivered to your inbox, or contact us.

Please take a few minutes to share insights regarding your experience with the AWS Public Sector Blog in this survey, and we’ll use feedback from the survey to create more content aligned with the preferences of our readers.

Scott Ma

Scott Ma

Scott Ma is a senior solutions architect at Amazon Web Services (AWS) based out of Florida. He specializes in architecting and building cloud native applications that enable customers to use best practices in their cloud journey. He is a builder at heart, with a passion for security and machine learning. In his spare time, he enjoys traveling, boating, and spending time with family and friends.