Desarrollode un simulador de redes de procesadores que evolucionan (NEPS) en la nube (SPARK)

Abstract

The natural-inspired computing has becomeone of the most frequently used techniques to handle complex problems such as the NP-Hard optimization problems. This kind of computing has several advantages over traditional computing, including resiliency, parallel data processing, and low consumptionof power. One of the active research areas of the natural-inspired algorithms is Network of Evolutionary Processors (NEPs). A NEP consists of several cells that are attached together; at the same time the edges of the graph are to transfer data between the nodes in system, while cells are representing the nodes.In this thesis we construct a NEPs system which is implemented over the Hadoop spark environment. The use of the spark platform is essential in this work due to the capabilities supplied by this platform. It is a suitable environment used solving some complicated problems. Using the environment is a possible choice in order to design the NEPs system. For this reason, in this thesis, we detailed on how to install, design and operate this system on the Apache the spark environment is used because it has the capability to implement the NEPs system in a distributed manner. The NEPs simulation is delivered in this work. An analysis of system’s parameters was also provided in this work for the system performance evaluation via the examination of each single factor affecting the performance of the NEPs individually. After testing the system, it become clear that using NEPs on the decentralize cloud eco-system can be thought as an effective method to handle data of different formats and also to execute optimization problems such as Adelman, 3-colorabilty and Massive-NEP problems. Moreover, this scheme is also robust that can be adaptable to handle data which might be scaled up to be big data which is characterized by its volume and heterogeneity. In this context heterogeneity might be referring to collecting data from different sources. Moreover, the utilization of the spark environment as a platform to operate the NEPs system has it is prospects. This environment is characterized by its fast task handing chunks of data to Hadoop architecture that is used to implement the spark system which is mainly based on the map and reduce functions. Thus, the task is distributed on NEPs system using the cloud based environment system made it possible to have logical result in all of the three examples investigated and examined in this method.