Parallel Mining of Correlated Heavy Hitters on Distributed and Shared-Memory Architectures

We present parallel algorithms for mining Correlated Heavy Hitters from a two-dimensional data stream. In particular, we design and implement a message-passing, a shared-memory and a hybrid algorithm. To the best of our knowledge, these are the first parallel algorithms solving the problem. We show, through experimental results, that our algorithms provide very good scalability, whilst retaining the accuracy of their sequential counterpart.