Data Modeling for a NoSQL Database Service

Cloud-hosted NoSQL database services, such as AWS DynamoDB, offer significant advantages, including low start-up costs, high performance and availability, wide scalability, and ease of deployment and management. These advantages have led to their rapid adoption and growth. However, the data storage, querying, and modification features supported by such NoSQL services are very rudimentary in comparison with those of relational and object database systems. Further, data modeling decisions made to map application requirements to the supported NoSQL model have very significant impact on not only performance but also financial cost incurred in using the services. Unlike the well developed body of work for relational-and object- database design, there is a great dearth of systematic procedures for NoSQL database design. This paper addresses this design problem by providing methods that map standard data models to the typically idiosyncratic and rudimentary models supported by NoSQL database services, using AWS DynamoDB as a specific instance.