Concurrent treaps

We propose algorithms to perform operations concurrently on treaps in a shared memory multi-core and multi-processor environment. Concurrent treaps hold the advantage of nodes’ priority for maintaining height of treaps. Concurrent treaps make use of logical ordering and physical ordering of nodes’ keys, and pessimistic locking mechanism to achieve synchronization. We observe that our concurrent treap implementations scale well as compared to the state-of-the-art implementations. We also study the impact of different locking objects on throughput of concurrent treaps. Our experimental results show that the concurrent treap implementation that uses AtomicInteger locking object provides better throughput and utilizes less memory footprint.