A Structure for Sliding Window Equijoins in Data Stream Processing

Sliding window equijoins are commonly used in data stream applications. In their implementation, a hash table is generally allocated for each stream source. However, this structure may degrade join performance because all tuples in the hash tables need to be scanned to find expired tuples whenever windows are updated periodically. This paper presents an alternative structure to improve the performance of sliding window equijoins. The basic idea of the proposed method is to allocate a hash table for each set of tuples arriving for a window slide interval. This organization facilitates the identification of expired tuples and improves the join performance, which we observed through our experimental results.