Ranking and benchmarking framework for sampling algorithmson synthetic data streams

In data streaming we work with large data from multiple sources. We observe overloaded partitions due to simple hash partitioning. We use partitioning algorithms that require a sampling algorithm to estimate the distribution of the stream. There is no standardized way to test these algorithms. We offer a framework with benchmark and hyperparameter optimization. Our work includes a generator for bursts and concept drifts. We provide algorithms that react to concept drifts