Relational (SQL) or NoSQL?
Today's applications have more demanding requirements than ever before. For example, an online game might start out with just a few users and a very small amount of data. However, if the game becomes successful, it can easily outstrip the resources of the underlying database management system. It is common for web-based applications to have hundreds, thousands, or millions of concurrent users, with terabytes or more of new data generated per day. Databases for such applications must handle tens (or hundreds) of thousands of reads and writes per second.
Amazon DynamoDB is well-suited for these kinds of workloads. As a developer, you can start small and gradually increase your utilization as your application becomes more popular. DynamoDB scales seamlessly to handle very large amounts of data and very large numbers of users.
For more information on traditional relational database modeling and how to adapt it for DynamoDB, see Best practices for modeling relational data in DynamoDB.
The following table shows some high-level differences between a relational database management system (RDBMS) and DynamoDB.
| Characteristic | Relational database management system (RDBMS) | Amazon DynamoDB |
|---|---|---|
| Optimal Workloads | Ad hoc queries; data warehousing; OLAP (online analytical processing). | Web-scale applications, including social networks, gaming, media sharing, and Internet of Things (IoT). |
| Data Model | The relational model requires a well-defined schema, where data is normalized into tables, rows, and columns. In addition, all of the relationships are defined among tables, columns, indexes, and other database elements. | DynamoDB is schemaless. Every table must have a primary key to uniquely identify each data item, but there are no similar constraints on other non-key attributes. DynamoDB can manage structured or semistructured data, including JSON documents. |
| Data Access | SQL is the standard for storing and retrieving data. Relational databases offer a rich set of tools for simplifying the development of database-driven applications, but all of these tools use SQL. | You can use the Amazon Web Services Management Console, the Amazon CLI, or NoSQL WorkBench to work with DynamoDB and perform ad hoc tasks. PartiQL, a SQL-compatible query language, lets you select, insert, update, and delete data in DynamoDB. Applications can use the Amazon software development kits (SDKs) to work with DynamoDB using object-based, document-centric, or low-level interfaces. |
| Performance | Relational databases are optimized for storage, so performance generally depends on the disk subsystem. Developers and database administrators must optimize queries, indexes, and table structures in order to achieve peak performance. | DynamoDB is optimized for compute, so performance is mainly a function of the underlying hardware and network latency. As a managed service, DynamoDB insulates you and your applications from these implementation details, so that you can focus on designing and building robust, high-performance applications. |
| Scaling | It is easiest to scale up with faster hardware. It is also possible for database tables to span across multiple hosts in a distributed system, but this requires additional investment. Relational databases have maximum sizes for the number and size of files, which imposes upper limits on scalability. | DynamoDB is designed to scale out using distributed clusters of hardware. This design allows increased throughput without increased latency. Customers specify their throughput requirements, and DynamoDB allocates sufficient resources to meet those requirements. There are no upper limits on the number of items per table, nor the total size of that table. |