Against All Probabilities: A modeling paradigm for streaming applications that goes against common notions

Hardware and software design requires the right portion of skills and mental faculties. The design of a good system is an exercise in rational thinking, engineering, and art. The design process is further complicated when we aspire to build systems that exploit parallelism or are targeted to be deployed on architecturally diverse computing devices, FPGAs or GPUs to name just a few. The need to develop systems that can take advantage of computing devices beyond general purpose CPUs is real. There are several application domains and research efforts that will simply not be able to adequately perform or yield answers in a reasonable amount of time otherwise. Developing a mathematical model of a system is a key stepping stone to a high performance system, but often is absent from the design process due to the complexity of the model development. In this paper we offer an easy, yet solid approach to the development of such mathematical models. This adds a little bit more weight to engineering side of the design process in the form of a quantifiable method that enables designers to reason about their systems, identify bottlenecks, and gain vital information for performance improvements