A functional flow framework for cloud computing

Abstract

This thesis covers a basic framework to calculate the maximum computation rate of a set of functions over a network. These functions are broken down into a series of computations, which are distributed among nodes of the network, with the output sent to the terminal node. We analyze two models with different types of computation costs, a linear computation cost model and a maximum computation cost model. We show how computation distribution through the given network changes with different types of computation and communication limitations. This framework can also be used in cloud design, where a network of given complexity is designed to maximize computation rate for a given set of functions. We provide a greedy algorithm that provides one solution to this problem, and create simulations for each framework, and analyze the results.