Time and Location: 3/11/2020, Wed., 4 p.m., CS Conference Room 206

Speaker #1: Mr. Zachary Hansen

Title: Data-driven Modeling and Predictions for Over-heating Events in HPC Centers

Abstract: High performance computing and big data processing are increasingly critical tools for a wide array of fields. As such, ensuring the optimal performance of the systems providing these services is of great interest. During the timeframe from December of 2018 to November of 2019 we collected physical readings such as temperature and power consumption from Texas Tech University’s data center, the High Performance Computing Center (HPCC). During the course of this data collection period, we recorded 17 separate overheating events, some of which necessitated the emergency shutdown of the entire data center. In an effort to prevent such events – which cost researchers time and money – we explored the feasibility of using neural networks to identify imminent overheating events. We compared the performance of a Direct Neural Network (DNN) with that of a Long Short-Term Memory network (LSTM).