Adaptive Parallel Approaches for GPU-Based Reencryption Applied in Scalable Proxy Servers for Protected Video Distribution

Abstract

Proxy servers of video distribution systems must provide efficient memory management and also have to adopt protection mechanisms so that only allowed clients would have access to restricted video content. This article extends previous work in which the usage of a GPU-based reencryption digital rights management module using a parallel version of AES cipher was proposed and evaluated, based on the idea that different adaptive parallel approaches would provide higher throughput depending on the GPU architecture. Three mix-grained approaches are proposed, implemented (CUDA, Pthreads and OpenMP) and evaluated on multicore CPU and GPU systems. An empirical analysis is conducted based on a multicore server with a 448-cores state-of-the-art GPU in contrast to a multicore desktop with a low-cost 128 cores graphical unit. Preliminary results demonstrate that the adaptive parallel approach impacts significantly, up to 33%, throughput provided by the reencryption module.