Fully Managed NoSQL Database designed to be highly scalable and performant
Supports enabling geo-replication of data with a click of a button
Has very low latency “less than 10ms”
Can perform Analytics on data using Spark
Designed for:
- Highly Regional (Globally distributed DB)
- Highly Responsive & Low Latency
Cosmos DB automatically indexes all the columns in the data to allow faster queries
Time to Live (TTL) option can be setup in Cosmos DB to delete stale data automatically from the container after a fixed period of time
In a Azure Subscription only 50 Cosmos DB Accounts can be created
Each account can contain any amount of data and databases
Request Unit (RU): Unit that encapsulates the resources used by DB. 1 RU roughly equals 1kb item read from a container
Cosmos DB APIs
The API to use has to be configured on the Cosmos DB Account
- SQL (Core)
- Document Datastore (Cosmos DB Native Datastore)
- Allows to query data using SQL
- Recommended option when moving from other SQL databases
- MongoDB
- Document Datastore
- Data stored in BSON format
- Cassandra
- Columnar Datastore
- Query data using CQL
- API currently supports OLTP workloads
- Gremlin
- Graph Datastore
- API supports OLTP workloads
- API is based on Apache TinkerPop
- Table
- Key Value Datastore
- Overcomes limitations of Azure Tables
- PostgreSQL
- Relational Database
- Uses Citus Open Source to Scale Horizontally
| SQL | Cassandra | MongoDB | Gremlin | Tables | PostgreSQL | |
|---|---|---|---|---|---|---|
| Cosmos Database | Database | Keyspace | Database | Database | NA | Database |
| Cosmos Container | Container | Table | Collection | Graph | Table | Table |
| Cosmos Entity | Item | Row | Document | Node/ Edge | Item | Row |
