Spyros

Lately, there is a new upcoming trend for accelerating security protocols on Graphics Processor Units (GPUs). The in parallel processing that GPUs offer, in combination with the several cores with which GPUs are equipped, reveal a promising and quite attractive alternative. Of course, there is always the drawback of power consumption (110-300 Watts), which could be 2-4 times more than the one that the latest generations of CPUs report. Nevertheless, the performance benefits especially for expensive computations (such as RSA operations) yield a 9-10 times faster processing. Until now, most of the experiments were targeting cryptanalysis tasks, but in the future we might see security protocols to be ported in languages like CUDA or OpenCL. Here is a very interesting and recent investigation for Accelerating TLS/SSL on GPUs. The comparison betweenGPUs and CPUs is very impressive. Imagine that the authors did not even patched the OpenSSL and they only used a web proxy structure in order to off-load the workload on GPUs. Already the first signs for porting security stacks appeared: Running CyaSSL on a GPU. Graphics Processor Units are portable, competitively cheap and extremely powerful. There are no guarantees that this emerging approach will be adopted, but all the facts show a great value for performance reasons.