Choosing the best database for modern applications across multiple data centers involves evaluating various factors, including availability, consistency, and durability. The belief that active-active architectures require a multi-master database is a misconception, as these often compromise data consistency and durability through conflict resolution strategies. Instead, sharded databases, such as MongoDB, provide a more robust solution by dividing data into shards, each with a primary server that handles writes, ensuring consistency and local read/write capabilities. MongoDB's architecture supports active-active configurations with features like fast failover, retryable writes, and zone sharding, making it suitable for global deployments. It offers fine-grained controls for write durability and read consistency, including causal consistency for ensuring logical read operations. Consequently, MongoDB is adeptly positioned to meet the demands of distributed applications while maintaining high performance and reliability.