waitdie

Wait-die is a deadlock avoidance protocol used in database systems to manage concurrent access to shared data. It is a timestamp-based concurrency control scheme and is commonly discussed alongside the wound-wait protocol. In wait-die, each transaction receives a unique timestamp when it starts, which establishes a global order of transactions.

When a transaction requests a lock on a data item currently held by another transaction with a conflicting lock, the protocol applies a simple rule. If the requesting transaction’s timestamp is older than the holder’s, the requester waits for the lock to be released. If the requester is younger, it is aborted (dies) and must restart with a new timestamp. Aborted transactions are typically rolled back to the point before the lock request and resubmitted.

This approach guarantees that no deadlocks occur, because a waiting chain cannot form a cycle with decreasing ages. However, it can lead to starvation for younger transactions in high-contention workloads, as repeated aborts may prevent timely progress. Wait-die is commonly used with two-phase locking and can be implemented in both centralized and distributed database environments.

Compared with wound-wait, where older transactions may abort younger ones to allow the older to proceed, wait-die reverses the direction: older transactions wait, younger ones are aborted. Each policy has trade-offs regarding abort frequency, potential starvation, and recovery costs, depending on workload characteristics. Implementations require careful management of timestamps, abort and restart logic, and handling of long-running lock holders. The choice between wait-die and wound-wait is typically guided by the expected contention pattern and the cost of aborts versus waits.