Fast and approximate stream mining of quantiles and frequencies using graphics processors

Abstract

We present algorithms for fast quantile and frequency estimation in large data streams using graphics processors (GPUs). We exploit the high computation power and memory bandwidth of graphics processors and present a new sorting algorithm that performs rasterization operations on the GPUs. We use sorting as the main computational component for histogram approximation and construction of ε-approximate quantile and frequency summaries. Our algorithms for numerical statistics computation on data streams are deterministic, applicable to fixed or variable-sized sliding windows and use a limited memory footprint. We use GPU as a co-processor and minimize the data transmission between the CPU and GPU by taking into account the low bus bandwidth. We implemented our algorithms on a PC with a NVIDIA GeForce FX 6800 Ultra GPU and a 3.4 GHz Pentium IV CPU and applied them to large data streams consisting of more than 100 million values. We also compared the performance of our GPU-based algorithms with optimized implementations of prior CPU-based algorithms. Overall, our results demonstrate that the graphics processors available on a commodity computer system are efficient stream-processor and useful co-processors for mining data streams.