Search

Various scientific applications with different I/O characteristics are executed in HPC systems. However, underlying parallel file systems are unaware of these characteristics of applications, and using a single fixed file layout for all applications can degrade the performance of HPC systems. In this paper, we propose A2FL, an autonomous and adaptive file layout adjustment scheme that optimizes parallel file system configurations by analyzing the access pattern of the applications. The key steps of A2FL are as follows: (1) A2FL initially intercepts the I/O operations of the application, recording their access patterns in real-time. (2) The access patterns are then transformed into a graphical representation used for predicting I/O performance and providing adjustment recommendations. (3) A2FL autonomously adjusts the file layout based on the prediction results, delivering an optimal file layout within the parallel file system. Moreover, we propose A2FL-Compound which analyzes an access pattern by dividing it into smaller components to optimize the file layout in a fine-grained manner. Our evaluations demonstrate that A2FL significantly enhances I/O performance, with improvements of up to 65.9× compared to the default file layout.

Large community of high-energy physicists share their data all around world making it necessary to ship a large number of files over wide- area networks. Regional disk caches such as the Southern California Petabyte Scale Cache have been deployed to reduce the data access latency. We observe that about 94% of the requested data volume were served from this cache, without remote transfers, between Sep. 2022 and July 2023. In this paper, we show the predictability of the resource utilization by exploring the trends of recent cache usage. The time series based prediction is made with a machine learning approach and the prediction errors are small relative to the variation in the input data. This work would help understanding the characteristics of the resource utilization and plan for additional deployments of caches in the future.

Energy consumption in buildings accounts for a significant portion of the global energy use. Consequently, understanding building energy use is important. Data over the past decade show that the energy intensity (Joules/sqft) of commercial buildings has decreased. While some of the improvements (decrease in energy use) are easily measurable such as the use of more energy efficient lighting, impact of other modifications such as changes to the operation of the HVAC system or changes in the usage pattern of the building potentially due to external events are difficult to quantify. Simply comparing energy consumption prior and post change is not accurate as energy use is impacted by many factors including external weather conditions. In this paper, we present a case study to quantify the impact of external events on the energy consumption of a medium-sized office building. We adopt an approach based on counterfactual analysis. Towards this end, we first build two models based on Linear Regression and k-Nearest Neighbors to predict the daily energy use given different input features related to the weather. We determine the statistical features of the weather that are most predictive of energy use. We then use the models to determine a counterfactual baseline and thereby to accurately estimate the impact of the events. The results of the counterfactual analysis provide new insights on the impact of the events on energy consumption. The update to the building cooling system resulted in more energy savings than direct yearly comparison reveals. On the other hand, the tests of a MPC-based controller for the HVAC system saved less energy than determined by the direct yearly comparison. Finally, the results show that there no gains in terms of energy savings due to remote work during the COVID-19 pandemic. An increase in airflow setting in the HVAC system corroborates this finding and further validates the underlying model and the counterfactual analyses.

Entry into parenthood is a major disruptive event to travel behavior, and gender gaps in mobility choices are often widened during parenthood. The exact timing of gender gap formation and their long-term effects on different subpopulations are less studied in the literature. Leveraging a longitudinal dataset from the 2018 WholeTraveler Study, this paper examines the effects of parenthood on a diverse set of short- to long-term outcomes related to the three hierarchical domains of mobility biography: mode choice, vehicle ownership, spatial mobility, and career decisions. The progress of the effects is evaluated over a sequential set of parenting stages and differentiated across three subpopulations. We find that individuals classified as “Have-it-alls”, who start their careers, partner up, and have children concurrently and early, significantly increase their car uses two years prior to childbirth (“nesting period”), and they then relocate to less transit-accessible areas and consequently reduce their reliance on public transportation while they have children in the household. In contrast, individuals categorized as “Couples”, who start careers and partnerships early but delay parenthood, and “Singles”, who postpone partnership and parenthood, have less pronounced changes in travel behavior throughout the parenting stages. The cohort-level effects are found to be driven primarily by women, whose career development is on average more negatively impacted by parenting events than men, regardless of their life course trajectory. Early career decisions made by women upon entering parenthood contribute to gender gaps in mid- to longer-term mobility decisions, signifying the importance of early intervention.

Scientific computing heavily relies on data shared by the community, especially in distributed data-intensive applications. This research focuses on predicting slow connections that create bottlenecks in distributed workflows. In this study, we analyze network traffic logs collected between January 2021 and August 2022 at the National Energy Research Scientific Computing Center (NERSC). Based on the observed patterns, we define a set of features primarily based on history for identifying low-performing data transfers. Typically, there are far fewer slow connections on well-maintained networks, which creates difficulty in learning to identify these abnormally slow connections from the normal ones. We devise several stratified sampling techniques to address the class-imbalance challenge and study how they affect the machine learning approaches. Our tests show that a relatively simple technique that undersamples the normal cases to balance the number of samples in two classes (normal and slow) is very effective for model training. This model predicts slow connections with an F1 score of 0.926.

Large-scale high performance computing (HPC) systems typically consist of many thousands of CPUs and storage units used by hundreds to thousands of users simultaneously. Applications from large numbers of users have diverse characteristics, such as varying computation, communication, memory, and I/O intensity. A good understanding of the performance characteristics of each user application is important for job scheduling and resource provisioning. Among these performance characteristics, I/O performance is becoming increasingly important as data sizes rapidly increase and large-scale applications, such as simulation and model training, are widely adopted. However, predicting I/O performance is difficult because I/O systems are shared among all users and involve many layers of software and hardware stack, including the application, network interconnect, operating system, file system, and storage devices. Furthermore, updates to these layers and changes in system management policy can significantly alter the I/O behavior of applications and the entire system. To improve the prediction of the I/O performance on HPC systems, we propose integrating information from several different system logs and developing a regression-based approach to predict the I/O performance. Our proposed scheme can dynamically select the most relevant features from the log entries using various feature selection algorithms and scoring functions, and can automatically select the regression algorithm with the best accuracy for the prediction task. The evaluation results show that our proposed scheme can predict the write performance with up to 90% prediction accuracy and the read performance with up to 99% prediction accuracy using the real logs from the Cori supercomputer system at NERSC.

The most common source of transformer failure is in the insulation, and the most prevalent warning signal for insulation weakness is partial discharge (PD). Locating the positions of these partial discharges would help repair the transformer to prevent failures. This work investigates algorithms that could be deployed to locate the position of a PD event using data from ultra-high frequency (UHF) sensors inside the transformer. These algorithms typically proceed in two steps: first determining the signal arrival time, and then locating the position based on time differences. This paper reviews available methods for each task and then propose new algorithms: a convolutional iterative filter with thresholding (CIFT) to determine the signal arrival time and a reference table of travel times to resolve the source location. The effectiveness of these algorithms are tested with a set of laboratory-triggered PD events and two sets of simulated PD events inside transformers in production use. Tests show the new approach provides more accurate locations than the best-known data analysis algorithms, and the difference is particularly large, 3.7X, when the signal sources are far from sensors.

In this work, we have analyzed the input/output (I/O) activities of Cori, which is a high-performance computing system at the National Energy Research Scientific Computing Center at Lawrence Berkeley National Laboratory. Our analysis results indicate that most users do not adjust storage configurations but rather use the default settings. In addition, owing to the interference from many applications running simultaneously, the performance varies based on the system status. To configure file systems autonomously in complex environments, we developed DCA-IO, a dynamic distributed file system configuration adjustment algorithm that utilizes the system log information to adjust storage configurations automatically. Our scheme aims to improve the application performance and avoid interference from other applications without user intervention. Moreover, DCA-IO uses the existing system logs and does not require code modifications, an additional library, or user intervention. To demonstrate the effectiveness of DCA-IO, we performed experiments using I/O kernels of real applications in both an isolated small-sized Lustre environment and Cori. Our experimental results shows that our scheme can improve the performance of HPC applications by up to 263% with the default Lustre configuration.

Data access throughput is one of the key performance metrics in scientific computing, particularly for distributed data-intensive applications. While there has been a body of studies focusing on elephant connections that consume a significant fraction of network bandwidth, this study focuses on predicting slow connections that create bottlenecks in distributed workflows. In this study, we analyze network traffic logs collected between January 2019 and May 2021 at National Energy Research Scientific Computing Center (NERSC). Based on the observed patterns from this data collection, we define a set of features to be used for identifying low-performing data transfers. Through extensive feature engineering and feature selection, we identify a number of new features to significantly enhance the prediction performance. With these new features, even the relatively simple decision tree model could predict slow connections with a F1 score as high as 0.945.

HPC and distributed systems are the driving force for the advancement of many emerging technologies, such as exascale systems, quantum machines, terabit networking, 5G/6G wireless, and cloud/edge computing. The tasks of systems and network telemetry are a key element for effective operations and management of the advancement of many emerging systems and technologies, and require more scalable telemetry and analysis techniques for comprehensive monitoring and analysis. Various input sources such as end systems, switches, firewalls, intrusion sensors and the emerging network elements speaking with different syntax and semantics make organizing and incorporating the generated data challenging for the quantitative and qualitative analysis. This workshop looks for new approaches and methods at the intersection of HPC systems and data sciences to address these difficult challenges of emerging technologies from the diverse angles of systems/network performance, availability, reliability, and security.

Federated Learning (FL) with mobile computing and the Internet of Things (IoT) is an effective cooperative learning approach. However, several technical challenges still need to be addressed. For instance, dividing the training process among several devices may impact the performance of Machine Learning (ML) algorithms, often significantly degrading prediction accuracy compared to centralized learning. One of the primary reasons for such performance degradation is that each device can access only a small fraction of data (that it generates), which limits the efficacy of the local ML model constructed on that device. The performance degradation could be exacerbated when the participating devices produce different classes of events, which is known as the class balance problem. Moreover, if the participating devices are of different types, each device may never observe the same types of events, which leads to the device heterogeneity problem. In this study, we investigate how data augmentation can be applied to address these challenges and improving detection performance in an anomaly detection task using IoT datasets. Our extensive experimental results with three publicly accessible IoT datasets show the performance improvement of up to 22.9% with the approach of data augmentation, compared to the baseline (without relying on data augmentation). In particular, stratified random sampling and uniform random sampling show the best improvement in detection performance with only a modest increase in computation time, whereas the data augmentation scheme using Generative Adversarial Networks is the most time-consuming with limited performance benefits.

Neural networks are powerful solutions to many scientific applications; however, they usually require long model training time due to large training data sets or large model size. Research has been focused on developing numerical optimization algorithms and parallel processing to reduce the training time. In this work, we propose a multi-resolution strategy that can reduce the training time by training the model with the reduced-resolution data samples at the beginning and later switching to the original resolution data samples. This strategy is motivated by the observation that coarser versions of many applications can be solved faster than their denser counterparts, and the solution to a coarser problem could be used to initialize the solution to the denser problem. When applying the idea to neural network training, coarse data can have a similar effect on the learning curves at the early stage as the dense data but requires less time. Once the curves no longer improve significantly, our strategy switches to using the data in original resolution. The key in this process is the ability to generate multiple resolutions of a problem automatically, which could usually be done with scientific applications with spatial and temporal continuity. We use two real-world scientific applications, CosmoFlow and DeepCAM, to evaluate the proposed mixed-resolution training strategy. Our experiment results demonstrate that the proposed training strategy effectively reduces the end-to-end training time while achieving a comparable accuracy to that of the training only with the original data. While maintaining the same model accuracy, our multi-resolution training strategy reduces the end-to-end training time up to 30% and 23% for CosmoFlow and DeepCAM, respectively.

Many scientific applications have started using deep learning methods for their classification or regression problems. However, for data-intensive scientific applications, I/O performance can be the major performance bottleneck. In order to effectively solve important real-world problems using deep learning methods on High-Performance Computing (HPC) systems, it is essential to address the poor I/O performance issue in large-scale neural network training. In this paper, we propose an asynchronous I/O strategy that can be generally applied to deep learning applications. Our I/O strategy employs an I/O -dedicated thread per process, that performs I/O operations independently of the training progress. The I/O thread reads many training samples at once to reduce the total number of I/O operations per epoch. Given the fixed amount of training data, the fewer the I/O operations per epoch, the shorter the overall I/O time. The I/O operations are also overlapped with the computations using the double-buffering method. We evaluate our I/O strategy using two real-world scientific applications, CosmoFlow and Neuron-Inverter. Our experimental results demonstrate that the proposed I/O strategy significantly improves the scaling performance without affecting the regression performance.

High-performance computing (HPC) systems consist of thousands of compute nodes, storage systems and high-speed networks, providing multiple layers of I/O stack with high complexity. By adjusting the diverse configuration settings that HPC systems provide, the I/O performance of applications can be improved. However, it is challenging to identify the optimal configuration settings without a thorough knowledge of the system, as each of the different I/O characteristics of applications can be an important factor for parameter decision. In this paper, we use multiple machine learning approaches to perform an in-depth analysis on I/O behaviors of HPC applications and to search for the optimal configuration settings for jobs sharing similar I/O characteristics. Improved by maximum 0.07 R-squared score, our results in overall show that jobs run on the HPC systems can obtain the predicted I/O performance for different configuration parameters with a high accuracy, using the proposed machine learning-based prediction models.

Variable selection (also known as feature selection) is essential to optimize the learning complexity by prioritizing features, particularly for a massive, high-dimensional dataset like network traffic data. In reality, however, it is not an easy task to effectively perform the feature selection despite the availability of the existing selection techniques. From our initial experiments, we observed that the existing selection techniques produce different sets of features even under the same condition (e.g., a static size for the resulted set). In addition, individual selection techniques perform inconsistently, sometimes showing better performance but sometimes worse than others, thereby simply relying on one of them would be risky for building models using the selected features. More critically, it is demanding to automate the selection process, since it requires laborious efforts with intensive analysis by a group of experts otherwise. In this article, we explore challenges in the automated feature selection with the application of network anomaly detection. We first present our ensemble approach that benefits from the existing feature selection techniques by incorporating them, and one of the proposed ensemble techniques based on greedy search works highly consistently showing comparable results to the existing techniques. We also address the problem of when to stop to finalize the feature elimination process and present a set of methods designed to determine the number of features for the reduced feature set. Our experimental results conducted with two recent network datasets show that the identified feature sets by the presented ensemble and stopping methods consistently yield comparable performance with a smaller number of features to conventional selection techniques.

While federated learning (FL) has gained great attention for mobile and Internet of Things (IoT) computing with the benefits of scalable cooperative learning and privacy protection capabilities, there still exist a great deal of technical challenges to make it practically deployable. For instance, the distribution of the training process to a myriad of devices limits the classification performance of machine learning (ML) algorithms, often showing a significantly degraded accuracy compared to centralized learning. In this paper, we investigate the problem of performance limitation under FL and present the benefit of data augmentation with an application of anomaly detection using an IoT dataset. Our initial study reveals that one of the critical reasons for the performance degradation is that each device sees only a small fraction of data (that it generates), which limits the efficacy of the local ML model (constructed by the device). This becomes more critical if the data holds the class imbalance problem, observed not infrequently in practice (e.g., a small fraction of anomalies). Moreover, device heterogeneity with respect to data quantity is an open challenge in FL. Based on these observations, we examine the impact of data augmentation on detection performance in FL settings (both homogeneous and heterogeneous). Our experimental results show that even a simple random oversampling can improve detection performance with manageable learning complexity.

To understand behaviors of natural and man-made events, such as energy consumption of buildings, which accounts for 40% of energy uses in the US, we deploy automated monitoring devices to record periodic observations. However, such experimental and observation data often contains problems and irregularities that have to be cleaned up before analyses. Due to various conditions affecting sensor operations, the communication channels, recording steps, or the recording media, the recorded data might have missing values, errors, or anomalous values. An effective way to clean up these problems is to replace these missing values, errors and anomalous values with expected values, a process generally known as imputation. In this work, we survey commonly used missing value imputation techniques and compare their performance on a set of building monitoring data. To compare the different types of sensor measurements with widely varying characteristics, we use normalized root mean squared error (NRMSE) as the key metric for the effectiveness of the imputation methods. We additionally consider periodicity and run time when considering comparing methods. Through extensive testing, we find that for small gap sizes, up to 8 consecutive missing values, linear interpolation performs the best; for larger gaps stretching up to 48 consecutive missing values, K-nearest neighbors provides the most accurate imputations; for even larger gaps, more computational intensive methods, such as matrix factorization, achieve the smallest NRMSE. Additionally, we observe that these computationally intensive algorithms not only provide accurate imputations for large gaps, but are also more robust across all types of sensors.

As data sets from DOE user science facilities grow in both size and complexity there is an urgent need for new capabilities to transfer, analyze and manage the data underlying scientific discoveries. LBNL's Superfacility project brings together experimental and observational research instruments with computational and network facilities at the National Energy Research Scientific Computing Center (NERSC) and the Energy Sciences Network (ESnet) with the goal of enabling user science. Here, we report on recent innovations in the Superfacility project, including advanced data management, API-based automation, real-Time interactive user interfaces, and supported infrastructure for 'edge' services.

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.

With the growth of 5G wireless technologies and IoT, it become urgent to develop robust network security systems, such as intrusions detection systems (IDS) to keep the networks secure. These IDS systems need to detect unauthorized access and attacks in real-time. However, most of the modern IDS are built based on complex machine learning models that are time-consuming to train. In this work, we propose a methodology using the SHapley Additive exPlanations (SHAP) in combination with tree-based classifiers. SHAP can be used to select consistent and small feature subsets to reduce the execution time and improve classification accuracy. We demonstrate the proposed approach with the Aegean Wi-Fi Intrusion Dataset (AWID) dataset in a series of multi-class classification experiments. Among the four classes ("normal", "injection", "flooding"and "impersonation"), it is well-known that the class impersonation is hard to be classified accurately. Tests show that we can use about 10% of the initial feature set without reducing the overall prediction accuracy. With this reduced set of features, the training time could be reduced as much as a factor of four, while slightly improving the discriminating ability to identify impersonation instances. This study suggests that by reducing the number of features, the classification algorithms are able to focus on key trends that differentiates the "attacks"classes from the "normal"class. Using a reduces subset of features improves IDS's accuracy and performance. Also, SHAP dependence plots capture the relationship between individual features and the classification decision.

Machine Learning (ML) methods have been widely used in Intrusion Detection Systems (IDS). In particular, many botnet detection methods are based on ML. However, due to the fast-evolving nature of network security threats, it is necessary to frequently retrain the ML tools with up-to-date data, especially because data labeling takes a long time and requires a lot of effort, making it difficult to generate training data. We propose transfer learning as a more effective approach for botnet detection, as it can learn from well curated source data and transfer the knowledge to a target problem domain not seen before. We devise an approach that is effective regardless whether or not the data from the target domain is labeled. More specifically, we train a neural network with the Recurrrent Variation Autoencoder (RVAE) structure on the source data, and use RVAE to compute anomaly scores for data records from the target domain. In an evaluation of this transfer learning framework, we use CTU-13 dataset as a source domain and a fresh set of network monitoring data as a target domain. Tests show that the proposed transfer learning method is able to detect botnets better than semi-supervised learning method that was trained on the target domain data. The area under Receiver Operating Characteristic is 0.810 for transfer learning, and 0.779 for directly using RVAE on the target domain data.

Large high-performance computers (HPC) are expensive tools responsible for supporting thousands of scientific applications. However, it is not easy to determine the best set of configurations for workloads to best utilize the storage and I/O systems. Users typically use the default configurations provided by the system administrators, which typically results in poor performance. In an effort to identify application characteristics more important to I/O performance, we applied several machine learning techniques to characterize these applications. To identify the features that are most relevant to the I/O performance, we evaluate a number of different feature selection methods, e.g., Mutual information regression and F regression, and develop a novel feature selection method based on Min-max mutual information. These feature selection methods allow us to sift through a large set of the real-world workloads collected from NERSC's Cori supercomputer system, and identify the most important features. We employ a number of different clustering algorithms, including KMeans, Gaussian Mixture Model (GMM) and Ward linkage, and measure the cluster quality with Davies Boulder Index (DBI), Silhouette and a new Combined Score developed for this work. The cluster evaluation result shows that the test dataset could be best divided into three clusters, where cluster 1 contains mostly small jobs with operations on standard I/O units, cluster 2 consists of middle size parallel jobs dominated by read operations, and cluster 3 include large parallel jobs with heavy write operations. The cluster characteristics suggest that using parallel I/O library MPI IO and a large number of parallel cores are important to achieve high I/O throughput.

Large-scale high performance computing (HPC) systems typically consist of many thousands of CPUs and storage units, while used by hundreds to thousands of users at the same time. Applications from these large numbers of users have diverse characteristics, such as varying compute, communication, memory, and I/O intensiveness. A good understanding of the performance characteristics of each user application is important for job scheduling and resource provisioning. Among these performance characteristics, the I/O performance is difficult to predict because the I/O system software is complex, the I/O system is shared among all users, and the I/O operations also heavily rely on networking systems. To improve the prediction of the I/O performance on HPC systems, we propose to integrate information from a number of different system logs and develop a regression-based approach that dynamically selects the most relevant features from the most recent log entries, and automatically select the best regression algorithm for the prediction task. Evaluation results show that our proposed scheme can predict the I/O performance with up to 84% prediction accuracy in the case of the I/O-intensive applications using the logs from CORI supercomputer at NERSC.

Large scientific projects are increasing relying on analyses of data for their new discoveries; and a number of different data management systems have been developed to serve this scientific projects. In the work-in-progress paper, we describe an effort on understanding the data access patterns of one of these data management systems, dCache. This particular deployment of dCache acts as a disk cache in front of a large tape storage system primarily containing high-energy physics data. Based on the 15-month dCache logs, the cache is only accessing the tape system once for over 50 file requests, which indicates that it is effective as a disk cache. The on-disk files are repeated used, more than three times a day. We have also identified a number of unusual access patterns that are worth further investigation.

Automated network intrusion detection systems (NIDS) continuously monitor the network traffic to detect attacks or/and anomalies. These systems need to be able to detect attacks and alert network engineers in real-time. Therefore, modern NIDS are built using complex machine learning algorithms that require large training datasets and are time-consuming to train. The proposed work shows that machine learning algorithms from the RAPIDS cuML library on Graphics Processing Units (GPUs) can speed-up the training process on large scale datasets. This approach is able to reduce the training time while providing high accuracy and performance. We demonstrate the proposed approach on a large subset of data extracted from the Aegean Wi-Fi Intrusion Dataset (AWID). Multiple classification experiments were performed on both CPU and GPU. We achieve up to 65x acceleration of training several machine learning methods by moving most of the pipeline computations to the GPU and leveraging the new cuML library as well as the GPU version of the CatBoost library.

The volume of data moving through a network increases with new scientific experiments and simulations. Network bandwidth requirements also increase proportionally to deliver data within a certain time frame. We observe that a significant portion of the popular dataset is transferred multiple times to different users as well as to the same user for various reasons. In-network data caching for the shared data has shown to reduce the redundant data transfers and consequently save network traffic volume. In addition, overall application performance is expected to improve with in-network caching because access to the locally cached data results in lower latency. This paper shows how much data was shared over the study period, how much network traffic volume was consequently saved, and how much the temporary in-network caching increased the scientific application performance. It also analyzes data access patterns in applications and the impacts of caching nodes on the regional data repository. From the results, we observed that the network bandwidth demand was reduced by nearly a factor of 3 over the study period.

Quickly identifying astronomical transients from synoptic surveys is critical to many recent astrophysical discoveries. However, each of the data processing pipelines in these surveys contains dozens of stages with highly varying time and space requirements. Properly predicting the resources required to run these pipelines is critical for the allocation of computing resources and reducing the discovery response time. We propose a machine learning strategy for this prediction task and demonstrate its effectiveness using a set of timing measurements from the intermediate Palomar Transient Factory (iPTF) workflow. The proposed model utilizes the spatiotemporal correlation of astronomical images, where nearby patches of the sky (space) are likely to have a similar number of objects of interest and workflows executed in the recent past (time) are likely to use a similar amount of time because the machines and data storage systems are likely to be in similar states. We capture the relationship among these spatial and temporal features in a Bayesian network and study how they impact the prediction accuracy. This Bayesian network helps us to identify the most influential features for predictions. With proper features, our models achieve errors close to the random variance boundary within batches of images taken at the same time, which can be regarded as the intrinsic limit of prediction accuracy.

We consider data sources connected to a software defined network (SDN) with heterogeneous link access rates. Deadline-driven data transfer requests are made to a centralized network controller that schedules pacing rates of sources and meeting the request deadline has a pre-assigned value. The goal of the scheduler is to maximize the aggregate value. We design a scheduler (RL-Agent) based on Deep Deterministic Policy Gradient (DDPG). We compare our approach with three heuristics: (i) PFAIR, which shares the bottleneck capacity in proportion to the access rates, (ii) VDRatio, which prioritizes flows with high value-to-demand ratio, and (iii) VBEDF, which prioritizes flows with high value-to-deadline ratio. For equally valued requests and homogeneous access rates, PFAIR is the same as an idealized TCP algorithm, while VBEDF and VDRatio reduce to the Earliest Deadline First (EDF) and the Shortest Job First (SJF) algorithms, respectively. In this scenario, we show that RL-Agent performs significantly better than PFAIR and VDRatio and matches and in over-loaded scenarios out-performs VBEDF. When access rates are heterogeneous, we show that the RL-Agent performs as well as VBEDF even though the RL-Agent has no knowledge of the heterogeneity to start with. For the value maximization problems, we show that the RL-Agent out-performs the heuristics for both homogeneous and heterogeneous access networks. For the general case of heterogeneity with different values, the RL-Agent performs the best despite having no prior knowledge of the heterogeneity and the values, whereas the heuristics have full knowledge of the heterogeneity and VDRatio and VBEDF have partial knowledge of the values through the ratios of value to demand and value to deadline, respectively.

Large community of high-energy physicists share their data all around world making it necessary to ship a large number of files over wide- area networks. Regional disk caches such as the Southern California Petabyte Scale Cache have been deployed to reduce the data access latency. We observe that about 94% of the requested data volume were served from this cache, without remote transfers, between Sep. 2022 and July 2023. In this paper, we show the predictability of the resource utilization by exploring the trends of recent cache usage. The time series based prediction is made with a machine learning approach and the prediction errors are small relative to the variation in the input data. This work would help understanding the characteristics of the resource utilization and plan for additional deployments of caches in the future.

Data caches of various forms have been widely deployed in the context of commercial and research and education networks, but their common positioning at the Edge limits their utility from a network operator perspective. When deployed outside the network core, providers lack visibility to make decisions or apply traffic engineering based on data access patterns and caching node location. As an alternative, in-network caching provides a different type of content delivery network for scientific data infrastructure, supporting on-demand temporary caching service. We will describe the status of in-network caching nodes deployed within ESnet in support of the US CMS data federation. We will describe the container and networking architecture used to deploy data caches within ESnet, and update on the evolving tooling around service management lifecycle. An analysis of cache usage will also be provided along with an outlook for expanding the in-network cache footprint.

The storage management system dCache acts as a disk cache for high-energy physics (HEP) data from the US ATLAS community. Since its disk capacity is considerably smaller than the total volume of ATLAS data, a heuristic is needed to determine what data should be kept on disks. An effective heuristic would be to keep the data files that are expected to be heavily accessed in the near future. Through a careful study of access statistics, we find a few most popular datasets are accessed much more frequently than others, even though these popular datasets change over time. If we could predict the near-term popularity of datasets, we could pin the most popular ones in the disk cache to prevent their accidental removal and guarantee their availability. To predict a dataset popularity, we present several methods for forecasting the number of times a dataset will be accessed in the next day. Test results show that these methods could predict the next-day access counts of popular datasets reliably. This observation is confirmed with dCache logs from two separate time ranges.

Wi-Fi has become the wireless networking standard that allows short- to medium-range device to connect without wires. For the last 20 year, the Wi-Fi technology has so pervasive that most devices in use today are mobile and connect to the internet through Wi-Fi. Unlike wired network, a wireless network lacks a clear boundary, which leads to significant Wi-Fi network security concerns, especially because the current security measures are prone to several types of intrusion. To address this problem, machine learning and deep learning methods have been successfully developed to identify network attacks. However, collecting data to develop models is expensive and raises privacy concerns. The goal of this paper is to evaluate a federated learning approach that would alleviate such privacy concerns. This initial work on intrusion detection is performed in a simulated environment. Once proven feasible, this process would allow edge devices to collaboratively update global anomaly detection models, without sharing sensitive training data. On a set of tests with the AWID intrusion detection data set, we show that our federated approach is effective in terms of classification accuracy, computation cost, as well as communication cost.

The demands of increasingly large scientific application workflows lead to the need for more powerful supercomputers. As the scale of supercomputing systems have grown, the prediction of fault tolerance has become an increasingly critical area of study, since the prediction of system failures can improve performance by saving checkpoints in advance. We propose a real-time failure detection algorithm that adopts an event-based prediction model. The prediction model is a convolutional neural network that utilizes both traditional event attributes and additional spatio-temporal features. We present a case study using our proposed method with six years of reliability, availability, and serviceability event logs recorded by Mira, a Blue Gene/Q supercomputer at Argonne National Laboratory. In the case study, we have shown that our failure prediction model is not limited to predict the occurrence of failures in general. It is capable of accurately detecting specific types of critical failures such as coolant and power problems within reasonable lead time ranges. Our case study shows that the proposed method can achieve a F1 score of 0.56 for general failures, 0.97 for coolant failures, and 0.86 for power failures.

We consider a science network that runs applications requiring data transfers to be completed within a given deadline. The underlying network is a software defined network (SDN) that supports fine grain real-time network telemetry. Deadline-aware data transfer requests are made to a centralized network controller that schedules the flows by setting pacing rates of the deadline flows and metering the background traffic at the ingress routers. The goal of the scheduling algorithm is to maximize the number of flows that meet the deadline while maximizing the network utilization. In this paper, we develop a Reinforcement Learning (RL) agent based network controller and compare its performance with well-known heuristics. For a network consisting of a single bottleneck link, we show that the RL-agent based network controller performs as well as Earliest Deadline First (EDF), which is known to be optimal. We also show that the RL-agent performs significantly better than an idealized TCP protocol in which the bottleneck link capacity is equally shared among the competing flows. We also study the sensitivity of the RL-agent controller for different parameter settings and reward functions.

The goal of this work is to investigate the impact of missing values in clustering joint categorical social sequences. Identifying patterns in sociodemographic longitudinal data is important in a number of social science settings. However, performing analytical operations, such as clustering on life course trajectories, is challenging due to the categorical and multidimensional nature of the data, their mixed data types, and corruption by missing and inconsistent values. Data quality issues were investigated previously on single variable sequences. To understand their effects on multivariate sequence analysis, we employ a dataset of mixed data types and missing values, a dissimilarity measure designed for joint categorical sequence data, together with dimensionality reduction methodologies in a systematic design of sequence clustering experiments. Given the categorical nature of our data, we employ an “edit” distance using optimal matching. Because each data record has multiple variables of different types, we investigate the impact of mixing these variables in a single dissimilarity measure. Between variables with binary values and those with multiple nominal values, we find that the ability to overcome missing data problems is more difficult in the nominal domain than in the binary domain. Additionally, alignment of leading missing values can result in systematic biases in dissimilarity matrices and subsequently introduce both artificial clusters and unrealistic interpretations of associated data domains. We demonstrate the usage of t-distributed stochastic neighborhood embedding to visually guide mitigation of such biases by tuning the missing value substitution cost parameter or determining an optimal sequence span.