Costs - Serverless 3D Data Optimization Pipelines on AWS

Costs

You are responsible for the costs of the AWS services used while running this solution. The total cost for running this solution depends on the size and complexity of the input test model. For full details, refer to the pricing webpage for each AWS service you will be using in this solution.

Example cost

With the solution deployed in the US East (N. Virginia) Region and using the default configuration, a 41MB input step file that takes 15 seconds for the PiXYZ scenario processor to convert, and a m4.large instance, costs approximately $0.07.

AWS Batch manages the on-demand EC2 instances for the processing. For this cost example, a single m4.large instance, scaling in after 10m21s was used. The delayed scale-in allows for subsequent jobs to use the started EC2 instances, amortizing the cold-start times for the instances.

Note: There are additional negligible service costs associated, such as $0.0009 for S3 and $0.00013 for SQS).

Table 2 - Summary of Costs

PiXYZ Scenario Processor (15 sec @ $15/hr) $ 0.25
EC2 (10m21s m4.large @ $.1/hr) $ 0.01
Total $ 0.26

In this example we are using the on-demand pricing for the PiXYZ Scenario Processor. This enables customers to innovate quickly with low risk since there are no upfront costs. However, some customers prefer to have annual fixed costs. For this option, PiXYZ offers a yearly fixed cost subscription ($20,000 per year) which allows unlimited use for a single container instance. If purchased, parallel tasks would first be metered to the yearly subscription until they are all consumed, then fallback the hourly metering. For example, if one yearly subscription was purchase and AWS Batch parallelized 64 tasks, the first task would be run at no charge (other than the yearly subscription), but the other 63 tasks would be metered at the hourly rate.

The advantages of fixed costs and unlimited usage from the yearly subscription must be balanced against rapid turn-around times through parallelization. If you have a constant, non-urgent workload of serializable conversions that exceeds 1333.3 hours of conversions per year (the equivalent of $20,000 at the hourly rate), we recommend using the hourly pricing so that you pay for only what you use and can aggressively parallelize the workloads for faster turn-around times.