Cost

You are responsible for the cost of AWS services that are used while operating this solution. As of this revision, the monthly cost of operating this solution for a small group of 25 users with limited model training and evaluation demand is about $7.78 per user (or $194.37 total).

The cost of operating the solution is dependent on the number of users registered to the system, the number of models they train, the number of evaluations they run, and the number of race submissions they perform. The cost of the solution is also dependent on the amount of time each model is trained for and the underlying instance type that is used for these jobs. The solution is configured to use an instance of type ml.c5.4xlarge by default. However, you may choose to change this based on your desired performance profile by updating the appropriate field in the solution’s CloudFormation template or CDK project and deploying that configuration.

Operating cost is figured and expressed on a monthly basis. The estimates provided in this section are only accurate for use cases in which all training jobs, evaluations, and submissions occur within the same month. This is an important consideration to account for when planning usage over the course of a quarter, semester, or other longer timeframe.

We recommend creating a budget through AWS Cost Explorer to help manage costs. Prices are subject to change. For full details, refer to the pricing webpage for each AWS service used in this solution.

Compute usage formula

The training job compute usage for the examples in this section is figured using the following formula:

Total SageMaker cost = [(212.5 hours) * ($0.816/instance hour)] + ($1.40 storage cost) = $174.80

Example 1 (25 users)

You are planning a deployment of DeepRacer on AWS that will serve 25 users, with each user creating and training 3 models. Each user will then evaluate each of their models 3 times, before submitting them to a race. With a training time of 90 minutes, the cost will be $194.37, or $7.78 per user.

Cost summary

Example 2 (100 users)

You are planning a deployment of DeepRacer on AWS that will serve 100 users, with each user creating and training 3 models. Each user will then evaluate each of their models 3 times, before submitting them to a race. With a training time of 90 minutes, the cost will be $716.28, or $7.17 per user.

Cost summary

Example 3 (250 users)

You are planning a deployment of DeepRacer on AWS that will serve 250 users, with each user creating and training 3 models. Each user will then evaluate each of their models 3 times, before submitting them to a race. With a training time of 90 minutes, the cost will be $1,760.09, or $7.04 per user.

Cost summary