Study of Scalability

A system having Scalability is highly desirable since it promises a proportional performance boost subject to the resource increase provided. In this chapter we try to analyze the extent of scalability various cryptographic workloads can offer. We try to optimally map two eSTREAM (ECRYPT (2012) “eSTREAM: The European Network of Excellence in Cryptology (ECRYPT) Stream Cipher Project.” [1]) finalists stream ciphers, i.e., HC-128 (and HC-256) and Salsa20 on modern Graphics Processor Units (GPUs). On NVIDIAs GPUs we used CUDA programming framework to exploit their many-core architecture on which parallel homogeneous threads are executed in a Single Instruction Multiple Thread (SIMT) fashion. Many cryptographic algorithms, especially block ciphers due to their block wise operations, have reportedly gained remarkable performance speedups on GPUs.