Amigos Library Services created an Excel workbook (see link below) to serve as a basic calculator to illustrate approximate infrastructure costs for deploying the SimplyE Circulation Manager service to Amazon Web Services (AWS). AWS is used for deployment by NYPL and many of the other Library Simplified partners, so AWS is the preferred cloud platform, especially as it relates to finding help in the community if you have questions during deployment.
The actual costs one might incur vary widely depending on the various AWS services an organization may choose to use, as well as the organization's existing practices and strategies (there is more than one way to deploy most elements of the system). The costs illustrated in the Excel workbook assume deployment of the Circulation Manager as a standalone service. The configuration also shows deploying the Circulation Manager using the project's Docker images on vanilla EC2 instances rather than using, say, the AWS ECS service. The last major assumption is that the deployment is for a single library. If multiple libraries are deployed on a single Circulation Manager service, the felt cost for each library in the multi-tenant system will be reduced. Your "mileage" will definitely vary.
Feel free to download a copy and massage the components and values to suit your organization's needs. The Unit and Cost/Mo columns include VLOOKUP formulas, which pull data from the
AWS Pricing sheet, based on the value in the Service Type column. Values changed in the Service Type column must conform exactly to their names in the
AWS Pricing sheet in order for their data to be located and retrieved accurately.
Notes about the pricing values:
- The workbook uses Amazon's On-Demand pricing, which is the more expensive option for deploying EC2 and RDS instances. If your organization can make a 1- or 3-year commitment to particular compute instance types, discounts of roughly 30-40% are available when using Reserved Instances. So, costs can be reduced significantly.
- Secondly, the items involving usage (data transfer and load balancer units, mainly) are pretty much guesses. These values will vary, and typically grow larger, over time as more patrons use the service. This factor is somewhat true of compute instance sizes, as well. The costs illustrate a system where use has matured, rather than one initially deployed.
- The odd 334GB size of EBS volumes in large deployments is a suggested size to alleviate an I/O burst rate limit observed in database processing–higher EBS storage values result in higher I/O rate limits. Amazon docs suggests allocating 334GB for a 1000 IOPS base rate when operating a service with high database I/O requirements.
- At the time of this writing, recent code changes may reduce the performance needs for the scripting and database instances (as well as the higher I/O limit) when maintaining a large OverDrive collection. We'll keep watch and update the spreadsheet in three to four months if the reduction is verified in production.