Intelligent Virtual Interface for Improving Performance in HPC Centers by Modelling Users and their Satisfaction

This paper deals with the implementation of a new approach that aims to improve the performance of high performance computing (HPC) centers through an agent based virtual interface that models the user behaviour and satisfaction. This interface translates user requests and satisfaction criteria into what is really necessary and, thus, permits the optimization of the HPC center's scheduler taking into account predicted user satisfaction together with the objectives of the management in terms of resource usage and cost. The system is based on an evolutionary profiling agent based architecture where the different agents are evolved in real time to adapt to the different users. These agents cooperate with the scheduling mechanism providing resource usage estimations for the different tasks, as well as predicting the effect of possible strategies on user satisfaction. Some experiments are presented where this system interacts with a real job management system (SGE) and we show that modelling the user and taking user satisfaction into account helps to improve system performance