Optimizing performance in a Cassandra cluster involves several best practices, including:

• Proper data modeling to minimize the number of partitions accessed per query and to distribute data evenly across nodes.
• Tuning configuration settings such as compaction and caching to balance read and write performance.
• Monitoring cluster health and performance metrics using tools like nodetool and DataStax OpsCenter to identify bottlenecks and optimize resource utilization.

Cassandra ensures data consistency and durability through its distributed architecture and replication strategies. It uses a tunable consistency level to control how many replicas must respond for a read or write operation to be considered successful. Additionally, Cassandra employs a commit log and a distributed storage engine to ensure durability by persisting data to disk before acknowledging a write operation as successful.

When designing a data model for Cassandra, it's essential to consider factors such as data distribution, query patterns, and data access patterns. Data modeling in Cassandra involves denormalization and optimizing the data model for the most common queries. Key considerations include choosing the right partition key, clustering columns, and secondary indexes to optimize query performance and data distribution across the cluster.

Cassandra supports schema flexibility, allowing for schema changes without downtime. Schema changes can be made using the ALTER KEYSPACE and ALTER TABLE statements, and new columns can be added without affecting existing data. Data migration in Cassandra can be achieved through various methods, including using tools like Apache Spark or DataStax Bulk Loader, or by implementing custom migration scripts to copy and transform data between tables or clusters. Additionally, rolling upgrades can be performed to update the Cassandra software version while maintaining cluster availability.