CosmosDB container = MongoDB collection = SQL tables
CosmosBD stores documents into containers.
- Partition Key:
- Each item (document) in a Cosmos DB container is associated with a partition key.
- The partition key is a property chosen by the user, and it determines the logical partition to which the item belongs.
- Physical Partitions:
- Cosmos DB distributes the data across physical partitions to achieve scalability.
- Each physical partition is a unit of throughput and storage capacity.
- Partition Key Range:
- A partition key range is a contiguous range of partition key values.
- Each physical partition contains one or more partition key ranges.
- The partition key ranges collectively cover the entire range of possible partition key values for a given container.
- Routing:
- Cosmos DB uses a mechanism called “range partitioning” to route requests to the correct physical partition based on the partition key.
- The partition key ranges help in efficiently routing requests to the appropriate physical partition.
- Scalability:
- Partition key ranges play a vital role in achieving horizontal scalability.
- As the volume of data increases, Cosmos DB can split or merge partition key ranges to balance the load across physical partitions.
- Query Performance:
- Well-distributed and evenly sized partition key ranges contribute to optimal query performance.
- Poorly chosen partition key values can lead to “hot partitions,” where a single partition key range receives a disproportionate amount of traffic.
- Throughput:
- Throughput (Request Units per Second, or RU/s) is provisioned at the level of physical partitions.
- Adjustments to throughput impact all partition key ranges within a physical partition.
Autoscaling
When using autoscale, Cosmos DB only scales the RU/s to the maximum when the RU consumption is 100%, for a continuous period of time in a 5 second interval.
More: link
