?
An optimistic approach to handle out-of-order events within analytical stream processing
In recent years, there has been a growth in research and industrial solutions in the field of distributed stream processing. However, even state-of-the-art stream processing systems experience difficulties with out-of-order data arrival. The most common solution to this issue is buffering. Its main problem is the extra cost for blocking before each order-sensitive operation. The goal of this paper is to propose and evaluate an optimistic approach to handle out-of-order events. We introduce a method that is suitable for any stateful operation and needs a single buffer for the complete computational pipeline. Such technique requires extra network transfers and re-computations, but the experiments demonstrate that a prototype of our approach is able to create low overhead while ensuring the correct ordering.