I have to write a system that has several data processing steps, each varying in time complexity. I intend to make use of buffer queues and multithreading for each preprocessing step. I have limited number of processes (each process can be seen as a server), and memory, so the buffer queues cannot be infinite. This seems like a perfect setting to use queueing models.
However, queueing models assume markov properties for input data, as I recall they also assume the input to follow poisson distribution.
So my question is, is it okay to assume those to model the data pipeline using queueing models?