Exploring Parallelism for Automatic and Scattered Investing on P2P Lending Platform

The features of P2P lending are wide coverage, low threshold, relative simple procedure, easy to satisfy the borrowers' requirement. In order to spread credit risk for lenders, automatic and scattered investing become more and more popular nowadays. In this paper, we take advantage of the computing power of modern GPUs to accelerate scattered investing in large-scale lender scenarios. By abstracting the scattered investing as map-reduce operation, we extend two stages map-reduce to four stages according to our scattered investing business and exploit a novel parallel investing system based on GPUs. Experimental results show that we achieve a maximum speedup of 27.52 times.