You are invited to the seminar of teni krihelY a student – MSc Optimal architecture of Edge cloud computation elements in cloud system architecture
M.Sc. student of Professor Yoram Reich
This thesis deals with the optimization of computing elements in an edge cloud system; we took as a case study a smart transportation system. The thesis presents a method, translated into an algorithm and code in Matlab, which can basically serve any system. The algorithm finds the optimal locations of computing elements in the system. In order to use the algorithm, it is necessary to define the following items:
- The existing components in the system.
- The functions performed by each component and the functions constituting the benefit or the desired performance of the system.
- The computational capability of each node and the cost of each computing element in each node.
- The relationships between the functions.
- Scenarios the system is undergoing; specifically, the connections and disconnections in each scenario between the functions.
The final value of the system is defined as its benefit to stakeholders divided by its cost. Nevertheless, it can also be defined differently by another method; there are several options as stated in the thesis. The solution method can operate in two ways: (1) in case of manageable number of elements, it runs over all the optional architectures (locations) of the computing elements in the system or (2) is uses a genetic algorithm to find the best alternatives.
The algorithm returns the following items:
- For specific objective, the best architecture for the placement of computing elements in the system, represented as a binary vector where each vector element indicates a different node. 0 means that there is no computing element in the node and 1 means that there is a computing element.
- For a multi-objective case, the algorithm finds the Pareto front, meaning the set of architectures that are defined as "undominated" set, meaning the best set, in terms of benefit and cost.