How It Works
DynamoDB is no different from other document style databases, using Hasing and B trees to organize your data into indexable collections. At a very basic level, each data object is stored in their own files. Then, another file stores key-value pairs with all of the documents in this collection. Then, as this collection gets larger there is an index file for the colection of index files across all of your documents, leading to a tiered system to allow ultra-fast access to your data. At the point where the amount of data you hold is larger than the server can handle, a paradigm called charding allows data to be stored on multiple servers with no impact to the end user. If your indexes were from 1-100, then one server would be responsible for storing 1-50, and the other 51-100.