Evaluation of Deep Learning Models for Network Performance Prediction for Scientific Facilities

Abstract

Large data transfers are getting more critical with the increasing volume of data in scientific computing. While scientific facilities manage dedicated infrastructures to support large data transfers, predicting network performance based on the historical measurement would be essential for workflow scheduling and resource allocation in the facility. In this study, we empirically evaluate deep learning (DL) models with respect to the prediction accuracy of network performance for scientific facilities, using a two-month network communication log. This paper compares a set of DL models based on Artificial Neural Network (ANN), Convolutional Neural Network (CNN), Gated Recurrent Unit (GRU), and Long Short-Term Memory (LSTM), to predict average throughput as a means to estimate network performance, and shares the observations made from the extensive experiments with the results of prediction accuracy and timing complexity.

Leveraging History to Predict Infrequent Abnormal Transfers in Distributed Workflows

“…The classification mechanism consists of two phases: assigning binary classification labels for each time window (anomalous or not) and performing actual classification by constructing a supervised learning model using the assigned label information. Another study in [22] evaluated deep learning models, including multilayer perceptron (MLP), convolutional neural network (CNN), gated recurrent unit (GRU), and long short-term memory (LSTM), in order to predict network performance (aggregated throughput) for each time interval. While these studies focused on analyzing tstat data based on time windows, our study focuses on connection-level prediction.…”

Predicting Slow Network Transfers in Scientific Computing

“…The classification mechanism consists of two phases, the first phase assigning binary classification labels for each time window (either anomalous or not), and the second phase performing actual classification by constructing a supervised learning model using the assigned label information. Another study in [11] performed the evaluation of deep learning models, including Multilayer perceptron (MLP), Convolutional Neural Network (CNN), Gated Recurrent Unit (GRU), and Long Short-Term Memory (LSTM), in order to predict throughput for each time interval. While these studies focused on analyzing tstat data based on time windows, our study focuses on the connection-level prediction.…”