Transaction Management and Concurrency Control Questions

Covers transaction management and concurrency control in database and storage systems, focusing on the four foundational properties of transactions: atomicity, consistency, isolation, and durability. Candidates should understand the transaction lifecycle including begin, commit, and rollback semantics; how atomicity and durability are implemented through write ahead logging, checkpoints, and recovery; and how consistency is enforced via integrity constraints and application invariants. Describe isolation levels such as Read Uncommitted, Read Committed, Repeatable Read, and Serializable, and explain the concrete anomalies each level permits or prevents, including dirty reads, non repeatable reads, lost updates, and phantom reads, as well as the concept of serializability as a correctness criterion. Cover concurrency control mechanisms including locking protocols with modes and granularity at row, page, and table levels, lock escalation, two phase locking, deadlock detection and resolution strategies, optimistic versus pessimistic concurrency strategies, and multi version concurrency control and snapshot isolation. Discuss performance and scalability trade offs when selecting isolation levels or concurrency strategies, practical diagnostics and monitoring for lock contention and deadlocks, and distributed transaction considerations such as two phase commit coordination, global ordering and partial failures, eventual consistency models, and when it is acceptable to relax strict transactional guarantees for availability or latency in distributed systems.