Accelerating data compression using general purpose GPUs

Abstract

The amount of data has been exponentially rising over the years and data centers have invested heavily in research for solutions to efficiently transport and store the data. One of the fields that is crucial to solving this problem is compression of the generated data to reduce the amount of hardware needed for storage of said data. Multiple solutions have already been explored and proposed, and in this paper we experiment with GPU parallelization of compression algorithms to reduce the time required to compress data while maintaining the maximal possible compression ratio. The experiments in this paper explore the viability of using general purpose GPUs for accelerating at home data compression, with the possibility of scaling up to large data centers i.e. using GPU parallelization to reduce compression times.