Search

Author(s): Dobson, P; Dwivedi, D; Millstein, D; Krishnaswamy, N; Garcia, J; Kiran, M | Abstract: Geothermal energy has traditionally been viewed as a baseload energy source, but the rapid growth of intermittent renewable energy has led to a need for more flexibility in power generation to avoid mandatory curtailment imposed by grid operators. This study of curtailment at The Geysers provides insights into the magnitude, duration, frequency, temporal and spatial distribution, and potential causes of curtailment events between 2013 and 2018. Annual levels of curtailment range during this period from 9 to 47 GW h, representing 0.15 to 0.81 % of the net generation. Most curtailments occurred at the power plants connected to a lower capacity transmission line and may result from transmission constriction. There is a clear link between negative pricing and economic curtailment, especially when solar production is higher. Economic curtailment events tend to be only a few hours and vary in magnitude up to almost 300 MW, whereas transmission-related curtailment events can be up to several weeks in duration. It is likely that curtailment of geothermal power will be an increasing concern, and could be mitigated by flexible generation strategies and increases in energy storage. It is critical to know the nature of curtailment events so that flexible generation options can be assessed properly.

In this paper, we explore both statistical and deep learning approaches for multi-step predictions in WAN traffic traces. Estimating future traffic can help improve link usage and optimize bandwidth utilization. In this paper, we study real network traces from a real WAN research network. We use Fourier analysis to present variation among the traffic traces, extracting daily and weekly peak frequencies per trace. We also develop statistical time-series methods, ARIMA and Holt-Winters, and three LSTM-based approaches with various neural network architectures (Simple, Stacked and S2S LSTM), to forecast and compare the accuracies between them. With efforts to find a data-driven learning solution, we find that deep learning approaches can learn traffic patterns and provide more accurate predictions than ARIMA and Holt-Winters. Our results show that predictions are improved at an average of 70% or more. We further discuss the challenges of building these, their deployment and how these can help improve network utilization for future planning problems.

International audience; Wide area networks are built to have enough resilience and flexibility, such as offering many paths between multiple pairs of end-hosts. To prevent congestion, current practices involve numerous tweaking of routing tables to optimize path computation, such as flow diversion to alternate paths or load balancing. However, this process is slow, costly and require difficult online decision-making to learn appropriate settings, such as flow arrival rate, workload, and current network environment. Inspired by recent advances in AI to manage resources, we present DeepRoute, a model-less reinforcement learning approach that translates the path computation problem to a learning problem. Learning from the network environment, DeepRoute learns strategies to manage arriving elephant and mice flows to improve the average path utilization in the network. Comparing to other strategies such as prioritizing certain flows and random decisions, DeepRoute is shown to improve average network path utilization to 30% and potentially reduce possible congestion across the whole network. This paper presents results in simulation and also how DeepRoute can be demonstrated by a Mininet implementation.

As the numbers of internet users and connected devices continue to multiply, due to big data and Cloud applications, network traffic is growing at an exponential rate. WAN networks, in particular, are witnessing very large traffic spikes cause by large file transfers that last from a few minutes to hours on network links and there is a need to develop innovative ways in which flows can be managed in real-time.In this work, we develop a reinforcement learning approach, in particular Upper-Confidence Algorithm, to learn optimal paths and reroute traffic to improve network utilization. We present throughput and flow diversions using Mininet and demo the technique using Chameleon’s Testbed (Bring-Your-Own-Controller [BYOC] functionality). This work is initial implementation towards DeepRoute, which combines Deep reinforcement learning algorithms with SDN controllers to create and route traffic using deployed OpenFlow switches.

Network traffic behavior is noisy and random, making it difficult to find patterns and predict future behavior. In this paper, we develop statistical models that use multivariate data model, incorporating seasonality, peak frequencies, and link relationships to improve future predictions. Using Fourier Transforms to extract seasons and peak frequencies from individual traces, we perform seasonality tests and ARIMA measures to determine optimal parameters to use in our prediction model. We develop a SARIMA multivariate model using real network traces to show improved prediction accuracy with better RMSE and smaller confidence intervals when compared to univariate approaches.

Network traffic has been a critical issue that has attracted massive attention in network operations research and the industry. This paper tackles the need to understand traffic patterns across a h...