DynamoDB is the latest nosql-database-as-a-service offering by Amazon. The design of DynamoDB is based on Amazon's Dynamo key-value store which is internally used by Amazon for their e-commerce platform. DynamoDB offers two main advantages over SimpleDB (Amazon's initial NoSQL offering based on Google's BigTable):
- Unlike SimpleDB which has limits on request rate and total size (10GB) per domain, DynamoDB can theoretically be provisioned for infinite throughput capacity per table. This means that DynamoDB takes care of sharding your data.
- Second, DynamoDB offers faster and more predictable query performance by offering a restricted data model and utilizing solid-state drives for storage.
At first it may seem that DynamoDB is a direct replacement for SimpleDB for all use cases. However, you should look at the data model and query flexibility required for your application and select the data store that best suites your application's needs. Let's take a look at DynamoDB's data model.
Data Model:
Tables, Items and Attributes
DynamoDB's data model consists of tables, items and attributes. Each table is a collection of items and each item is a collection of attributes. An Attribute is simply a name-value pair, e.g., ("userId"=1234). Consider for example a Users table where a user item may look like:
Users
=====
"UserId"=1234
"Name"="Bilal Sheikh"
"City"="Waterloo"
...
...
Posts
=====
"PostId"=345
"Version"=2
...
...
Primary Key, Composite Key and Indices
Each table must have a primary key attribute which is specified when the table is created. The primary key attribute must have a value for all items and the value must be unique. In our User table, UserId is the primary key attribute.
Instead of specifying a single attribute as the primary key, a composite primary key can be specified on two attributes. Hash index used for partitioning data is created for one of the attributes and a range index in created for the second attribute of a composite key. In our Posts model, a composite primary key of UserId and DateTime can be specified where the UserId will be the hash key and the DateTime will be the range key. Both attributes of a composite key must have a value for all items.
Note: The data model is much more restricted than SimpleDB where indices are created automatically for all attributes. The restricted data model allows for theoretically infinite scalability and fast get and put operations.
Basic API and provisioning:
DynamoDB offers a basic API for table creation, deletion and setting the read/write provisioning for a table. Put, get, update and delete operations are supported for individual items. Updates can be conditional and can be used for optimistic concurrency control. DynamoDB also offers a query operation for searching based on the values of primary key attributes and a scan operation which as the name implies can be used to scan the entire table.
For each DynamoDb table we have to specify the read and the write provisioning. Read or write unit capacity is measured as the number of 1KB items read or written per second. For example, if we provision for 5 read units for a table, we can execute 100 read requests in a second where each request reads 50 bytes of data. Alternatively, only two requests can be executed if each request reads 2Kb of data. The read and write capacity limits the total bytes of data read or written per second. More detail regarding calculation of read and write units can be found here.