📊 Cosmos DB - Advanced

Global Distribution

from azure.cosmos import CosmosClient, PartitionKey

client = CosmosClient.from_connection_string(connection_string)

# Create geo-replicated database
database = client.create_database(
  id="mydb",
  offer_throughput=400,
  consistency_level="Session"
)

# Add read regions
client._client_connection._endpoint_properties.update({
  "multipleWriteLocationsEnabled": True,
  "writeLocations": [
    {"locationName": "East US", "failoverPriority": 0},
    {"locationName": "West US", "failoverPriority": 1}
  ]
})

Partition Strategy Optimization

def choose_partition_key(data_distribution):
  """
  Analyze data distribution to choose optimal partition key
  """
  if data_distribution['skew_ratio'] > 0.8:
    return "/userId"  # Hot partition risk
  else:
    return "/timestamp"  # Better distribution

Conflict Resolution

container = database.create_container(
  id="conflicts",
  partition_key=PartitionKey(path="/userId"),
  conflict_resolution_policy={
    "mode": "LastWriterWins",
    "conflictResolutionPath": "/_ts"
  }
)

Key Takeaways

• **Multi-region** writes enable global scalability

• **Partition keys** impact performance

• **Conflict resolution** handles simultaneous updates

• **Consistency levels** trade latency for guarantees

Next Steps

Learn about monitoring and observability.