Malachi Schram

Data provenance hybridization supporting extreme-scale scientific workflow applications

As high performance computing (HPC) infrastructures continue to grow in capability and complexity, so do the applications that they serve. HPC and distributed-area computing (DAC) (e.g. grid and cloud) users are looking increasingly toward workflow solutions to orchestrate their complex application coupling, pre- and post-processing needs. To that end, the US Department of Energy Integrated end-to-end Performance Prediction and Diagnosis for Extreme Scientific Workflows (IPPD) project is currently investigating an integrated approach to prediction and diagnosis of these extreme-scale scientific workflows. To gain insight and a more quantitative understanding of a workflows performance our method includes not only the capture of traditional provenance information, but also the capture and integration of system environment metrics helping to give context and explanation for a workflows execution. In this paper, we describe IPPDs provenance management solution (ProvEn) and its hybrid data store combining both of these data provenance perspectives. We discuss design and implementation details that include provenance disclosure, scalability, data integration, and a discussion on query and analysis capabilities. We also present use case examples for climate modeling and thermal modeling application domains.

Belle II production system

The Belle II experiment will record a similar quantity of data to LHC experiments and will acquire it at similar rates. This requires considerable computing, storage and network resources to handle not only data created by the experiment but also considerable amounts of simulated data. Consequently Belle II employs a distributed computing system to provide the resources coordinated by the the DIRAC interware. DIRAC is a general software framework that provides a unified interface among heterogeneous computing resources. In addition to the well proven DIRAC software stack, Belle II is developing its own extension called BelleDIRAC. BelleDIRAC provides a transparent user experience for the Belle II analysis framework (basf2) on various environments and gives access to file information managed by LFC and AMGA metadata catalog. By unifying DIRAC and BelleDIRAC functionalities, Belle II plans to operate an automated mass data processing framework named a production system. The Belle II production system enables large-scale raw data transfer from experimental site to raw data centers, followed by massive data processing, and smart data delivery to each remote site. The production system is also utilized for simulated data production and data analysis. Although development of the production system is still on-going, recently Belle II has prepared prototype version and evaluated it with a large scale simulated data production. In this presentation we will report the evaluation of the prototype system and future development plans.

Towards efficient scheduling of data intensive high energy physics workflows

Data intensive high energy physics workflows executed on geographically distributed resources pose a tremendous challenge for efficient use of computing resources. In this early work paper, we present a hierarchical framework for efficient allocation of resources and energy-efficient assignment of tasks for a representative high energy physics application, the Belle II experiments. With an expected data rate of 25 peta bytes per year from experimental data and Monte Carlo simulations, the Belle II experiment provides an ideal platform for algorithmic development. Building on the analogy of the unit commitment problem in electric power grids, we present a novel cost-efficient method for resource allocation that feeds into energy-efficient assignment of tasks to resources using a novel semi-matching based algorithm. We demonstrate that this approach is both computationally efficient and effective. We expect the methods developed in this work to benefit Belle II and other complex workflows executed on distributed resources.

Towards Efficient Resource Allocation for Distributed Workflows Under Demand Uncertainties

Scheduling of complex scientific workflows on geographically distributed resources is a challenging problem. Selection and scheduling of a subset of available resources to meet a given demand in a cost efficient manner is the first step of this complex process. In this paper, we develop a method to compute cost-efficient selection and scheduling of resources under demand uncertainties. Building on the techniques of Sample Average Approximation and Genetic Algorithms, we demonstrate that our method can lead up to (24%) improvement in costs when demand uncertainties are explicitly considered. We present the results from our preliminary work in the context of a high energy physics application, the Belle II experiments, and believe that the work will equally benefit other scientific workflows executed on distributed resources with demand uncertainties. The proposed method can also be extended to include uncertainties related to resource availability and network performance.

Leveraging large sensor streams for robust cloud control

Todays dynamic computing deployment for commercial and scientific applications is propelling us to an era where minor inefficiencies can snowball into significant performance and operational bottlenecks. Data center operations is increasingly relying on sensors based control systems for key decision insights. The increased sampling frequencies, cheaper storage costs and prolific deployment of sensors is producing massive volumes of operational data. However, there is a lag between rapid development of analytical techniques and its widespread practical deployment. We present empirical evidence of the potential carried by analytical techniques for operations management in computing and data centers. Using Machine Learning modeling techniques on data from a real instrumented cluster, we demonstrate that predictive modeling on operational sensor data can directly reduce systems operations monitoring costs and improve system reliability.

Method of Fission Product Beta Spectra Measurements for Predicting Reactor Anti-neutrino Emission

The nuclear fission process that occurs in the core of nuclear reactors results in unstable, neutron-rich fission products that subsequently beta decay and emit electron antineutrinos. These reactor neutrinos have served neutrino physics research from the initial discovery of the neutrino to todays precision measurements of neutrino mixing angles. The prediction of the absolute flux and energy spectrum of the emitted reactor neutrinos hinges upon a series of seminal papers based on measurements performed in the 1970s and 1980s. The steadily improving reactor neutrino measurement techniques and recent reconsiderations of the agreement between the predicted and observed reactor neutrino flux motivates revisiting the underlying beta spectra measurements. A method is proposed to use an accelerator proton beam delivered to an engineered target to yield a neutron field tailored to reproduce the neutron energy spectrum present in the core of an operating nuclear reactor. Foils of the primary reactor fissionable isotopes placed in this tailored neutron flux will ultimately emit beta particles from the resultant fission products. Measurement of these beta particles in a time projection chamber with a perpendicular magnetic field provides a distinctive set of systematic considerations for comparison to the original seminal beta spectra measurements. Ancillary measurements suchmorexa0» as gamma-ray emission and post-irradiation radiochemical analysis will further constrain the absolute normalization of beta emissions per fission. The requirements for unfolding the beta spectra measured with this method into a predicted reactor neutrino spectrum are explored.«xa0less

Deep Learning on Operational Facility Data Related to Large-Scale Distributed Area Scientific Workflows

Distributed computing platforms provide a robust mechanism to perform large-scale computations by splitting the task and data among multiple locations, possibly located thousands of miles apart geographically. Although such distribution of resources can lead to benefits, it also comes with its associated problems such as rampant duplication of file transfers increasing congestion, long job completion times, unexpected site crashing, suboptimal data transfer rates, unpredictable reliability in a time range, and suboptimal usage of storage elements. In addition, each sub-system becomes a potential failure node that can trigger system wide disruptions. In this vision paper, we outline our approach to leveraging Deep Learning algorithms to discover solutions to unique problems that arise in a system with computational infrastructure that is spread over a wide area. The presented vision, motivated by a real scientific use case from Belle II experiments, is to develop multilayer neural networks to tackle forecasting, anomaly detection and optimization challenges in a complex and distributed data movement environment. Through this vision based on Deep Learning principles, we aim to achieve reduced congestion events, faster file transfer rates, and enhanced site reliability.