San Diego Supercomputer Center Research Articles

A DFT+U Study of the Electrochemical Oxidation of H2 and CO on SrLaFeO4-Δ

INTRODUCTION Solid oxide fuel cell (SOFC) anodes are commonly composed of cermet materials, typically Ni-YSZ. Despite success in electrochemical performance, cermet materials suffer from sulfur poisoning, coke formation under low operating temperatures, and metal coarsening. Due to success of material design focused on the cathode such as LSCF, single-phase mixed ionic-electronic conducting (MIEC) materials have displayed promising oxidation activity [1]. The Sr-based Ruddlesden-Popper (RP) family has found recent use as a base stoichiometry in anodes where experimentalists have used both A and B-site doping and nanoparticle exsolvation strategies in order to tune activity [2,3]. In order to understand the fundamental electrochemical activity, we want to elaborate on the mechanism of H2 and CO oxidation on the promising material SrLaFeO4- \U0001d6ff (SLF). In the present study, we used ab initio methods to model H2 and CO oxidation on two FeO2-based models of SLF with and without B-site doping with Co and Ni. COMPUTATIONAL DETAILS All calculations presented in this work were performed using the spin-polarized DFT+U method with periodic boundary conditions as implemented in the VASP 5.4.4 code. We choose the PBE functional to describe exchange and correlation effects. Dudarev’s approach for DFT+U calculations is used with a U-J value of 4.0, 3.32, and 6.2 eV, respectively. We used the projector-augmented wave (PAW) method to represent the inner core potentials. The kinetic energy cutoff was set for all calculations to 700 eV, and integration over the Brillouin zone was performed with 3 × 3 × 1 Monkhorst-Point k-point mesh and the Gaussian smearing method with a sigma value of 0.05 eV. Slab models utilized a 15 Å vacuum layer and the bottom two layers were fixed to bulk position values. The DFT-derived parameters were employed in a microkinetic model to calculate the surface coverage of all adsorbed species for a given reaction condition. RESULTS AND DISCUSSION To begin our study, we tested all non-symmetrical conformers of SLF for the lowest relative energy on a model 2 × 2 × 1 supercell, as displayed in Figure 1a. A total of 3 structures were tested – P4/mmm, P4/nmm, and I4mm. We confirm that the lowest energy structure (I4mm) is equivalent to the experimentally reported structure. Next, we develop two 1.5 × 1.5 × 1 non-identical (001) FeO2-based slab models – so-called FeO2-LaO and FeO2-SrO (Figure 2b). The oxidation mechanism of H2 suggests oxidation begins with a dissociative adsorption step. Subsequent steps include formation of surface H2O, a surface vacancy, and a subsurface vacancy. To complete the catalytic cycle, we computed the bulk vacancy migration of SLF and assumed fast cathode kinetics for the overall cell model. At an operating voltage of 0.7 V, we predict that the highest rate occurs on the FeO2-SrO surface. We predict that H2 dissociative adsorption step is the rate-determining process based on Campbell’s degree of rate control. With the application of a single B-site dopant (Co and Ni), Co and Ni both decrease the barrier of the H2 dissociative adsorption step. Ni causes a slight decrease in energy for the formation energy of H2O. The oxidation mechanism of CO suggests oxidation begins with adsorption of the CO molecule on a reduced Fe site. The CO migrates to a neighboring Fe-O complex in order to form surface CO2. Based on degree of rate control, CO2 formation is likely the rate-determining step for both surface models.In conclusion, this is one of the first mechanism investigations on the anodic oxidation reactions of H2 and CO on RP-based oxides. We find that H2 has promising catalytic performance similar to the experimental evidence. The inclusion of B-site dopants displays that catalytic performance can be tuned with the introduction of different transition metals. Future work will examine the role of exsolved particles for RP-based oxides. ACKNOWLEDGMENTS This work has been funded by the Division of Materials Research, a National Science Foundation organization under Award Number 1832809. Computing resources provided by the San Diego Supercomputer Center (SDSC) and Texas Advanced Computing Center (TACC) under XSEDE grant number TG-CTS090100 and USC High-Performance Computing clusters are gratefully acknowledged. REFERENCES L. Xu, Y. Yin, N. Wang, Z. Ma, Electrochem. Commun., 76, 51–54 (2017)N. Wu, W. Wang, Y. Zhong, G. Yang, J. Qu, Z. Shao, ChemElectroChem, 4, 2378–2384 (2017)H. Chong, H. Chen, Z. Shao, J. Shi, J. Bai, S. Li, J. Mater. Chem. A, 4, 13997–14007 (2016) FIGURE 1a: 2×2×1 supercell model of I4mm SrLaFeO4 FIGURE 1b: Slab models of (001) FeO2-LaO and FeO2-SrO Figure 1

Performance evaluation of an adopted model based on big-bang big-crunch and artificial neural network for cloud applications

High-performance computing is changing the way we compute. In the past decade, the cloud computing paradigm has changed the way we compute, communicate, and technology. Cover real-world problems. There are still many complex challenges in the cloud computing paradigm. Improving effective planning strategies is a complex problem in the service-oriented computing paradigm.In this article, our research focuses on improving task scheduler strategies to improve the performance of cloud applications. The proposed model is inspired by an artificial neural network-based system and astrology base scheduler Big-Bang Big-Crunch. The results show that the proposed strategy based on BBBC and neural network is superior to the method based on astrology (BigBang BigCrunch costaware), genetic cost and many other existing methods.The proposed BB-BC-ANN model is validated using standard workload file (San Diego Supercomputer Center (SDSC) Blue Horizon logs). The results show that the proposed BB-BC-ANN model performs better than some of the existing approaches using performance indicators like total completion time (ms), average start time (ms), average finish time(ms), scheduling time(ms), and total execution time(ms).

A DFT+U Study of the Oxygen Defect Formation in (Sr1-xPrx)2FeO4

INTRODUCTION Solid oxide fuel cell (SOFC) electrodes are commonly composed of cermet materials, a composite of ceramic and metal materials. Despite success in electrochemical performance, cermet materials suffer from sulfur poisoning at the triple-phase boundary (TPB), high operating temperatures, and metal coarsening. To mitigate these problems, single-phase mixed ionic-electronic conducting (MIEC) materials have been investigated as alternative electrode materials. One such class of materials is the Ruddlesden-Popper (RP) phase (An+1BnO3n+1), a perovskite composed of n layers of octahedra in a perovskite (ABO3) stack. The RP family of Srn+1FenO3n+1 has found recent use as a base stoichiometry in both cathodes1 and anodes2 with and without A-site trivalent dopants. In order to rationalize the design of RP electrodes, we elaborate on the relationship of MIEC properties as a function of A-site dopant concentration at the atomistic level. In the present study, we used ab initio methods to investigate the relationship between Pr3+ concentration, iron oxidation state, and charge compensation upon defect formation to better understand the role of the A-site on the ionic and electronic properties in (Sr1-xPrx)2FeO4± \U0001d6ff (x = 0, 0.125, 0.25, 0.375, 0.5). COMPUTATIONAL DETAILS All calculations presented in this work were performed using the spin-polarized DFT+U method with periodic boundary conditions as implemented in the VASP 5.4.4 code. We choose the PBE functional to describe exchange and correlation effects. Dudarev’s approach for DFT+U calculations is used to correct the inadequate description of localized 3d electrons on Fe and 4f electrons on Pr with a U-J value of 4.0 and 6.0 eV, respectively. We used the projector-augmented wave (PAW) method to represent the inner core potentials. The kinetic energy cutoff was set for all calculations to 750 eV, and integration over the Brillouin zone was performed with 5×5×3 Monkhorst-Point k-point mesh and the tetrahedron method with Blöchl corrections. RESULTS AND DISCUSSION To sample all possible structural conformations at a given dopant concentration, we tested all non-symmetrical conformers for the lowest relative energy on a model 2×2×1 supercell, as displayed in Figure 1a. A total of 7, 42, 123, and 181 structures were tested for each dopant configuration, respectively. We calculate metallic behavior for x = 0.125 and 0.25 and semi-conducting behavior for x = 0, 0.375 and 0.5, indicating electronic conductivity decreases upon increasing A-site dopant concentration. Figure 1b displays the relationship between the final optimized dopant configurations and defect stability. The interstitial formation energy decreases linearly with Pr concentration with a deviation at x = 0.125. This relationship is primarily a function of the neighboring cation elemental identity and the decreasing interatomic distance between the interstitial defect and the neighboring cations in the rocksalt regime. The vacancy formation energy appears constant with Pr except for a large energy penalty at x = 0.5. The origin of the harsh energy penalty is rooted in the over-reduction of Fe. Upon vacancy formation, the released electrons directly reduce the next-nearest Fe atoms from +3 to +2 as opposed to delocalizing upon the oxide lattice. Figure 1c displays the relationship between the degree of delocalization and the energy of vacancy formation. The degree of delocalization3 is a dimensionless quantity that describes the extent of charge delocalized upon the oxygen sublattice with a vacancy. For x = 0 to 0.375, vacancy formation delocalizes the released electrons onto the lattice as opposed to x = 0.5, where charge from the lattice directly reduces Fe. Covalence between Fe-O bonds is decreased upon the formation of Fe2+, causing a harsh penalty in vacancy formation.In conclusion, tuning MIEC behavior is a balance between reducing Fe and charge delocalization upon the oxide lattice. In terms of electronic conductivity, a small introduction of Pr yielded metallic behavior. In terms of ionic conductivity, a threshold value of Pr yields an unfavorable vacancy formation energy. Future studies will focus on the optimization of the catalytic activity by B-site doping. ACKNOWLEDGMENTS This work has been funded by the Division of Materials Research, a National Science Foundation organization under Award Number 1832809. Computing resources provided by the San Diego Supercomputer Center (SDSC) and Texas Advanced Computing Center (TACC) under XSEDE grant number TG-CTS090100 and USC High-Performance Computing clusters are gratefully acknowledged. REFERENCES T. Hong, M. Zhao, K. Brinkman, F. Chen, C. Xia, ACS Appl. Mater. Interfaces, 9, 8659–8668 (2017)C. Yang, Z. Yang, C. Jin, G. Xiao, F. Chen, Adv. Mater., 11, 1439-1443 (2012)A. M. Ritzmann, J. M. Dieterich, E. A. Carter, Phys. Chem. Chem. Phys., 17, 12260-12269 (2016) Figure 1

Power efficient resource provisioning for cloud infrastructure using bio-inspired artificial neural network model

High-performance computing changes the way of computing. More than one-decade, the cloud computing paradigm has changed the way of computing, communication, and technology. The efficient resource provisioning or task scheduler policy improvement is a challenging issue in the service-oriented computing paradigm. This article work focuses on task scheduler policy improvement for better cloud application performance. The task scheduling algorithm aims to improve the performance of real-time applications in the cloud by reducing task waiting time, execution time, and power consumption. The proposed model is inspired by an Artificial Neural Network (ANN) based system with a training model using a genetic algorithm. Results exhibit that the proposed GA-ANN policy outperforms the (Big-Bang Big-Crunch cost-aware), Genetic cost-aware, and other existing approaches. The results show that the proposed GA-ANN model performs better than existing approaches taking power consumption, total completion time(ms), average start time (ms), and average completion time(ms) as performance metrics. The proposed GA-ANN model is validated using real-time user requests (standard workload file) from workload traces (San Diego Supercomputer Center (SDSC) Blue Horizon logs), and fabricated data sets. The proposed model improves power efficiency by 13 %, scheduling time by 77.14 % and total execution time by 36 %. Hence the proposed GA-ANN technique provides performance as compared to existing systems.

Abstract IA20: California Teachers Study (CTS) Data Management Platform: A model for a repeatable turnkey, end-to-end, cloud-based data management and analytics solution for epidemiology cohorts

Abstract The Health Cyberinfrastructure (CI) Division at the San Diego Supercomputer Center (SDSC) has been deploying secure, compliant, end-to-end solutions to support critical applications including, but not limited to, biomedical research, detecting and preventing medical fraud, enterprise risk management, and medical device data management. These applications have relied on technologies ranging from the traditional data warehouse to the more recent big data platforms leveraging Hadoop, Software as a Service (SaaS) capability leveraging container technology, business intelligence and analytics solutions, and IOT/streaming solutions. Operating these applications and the underlying platforms requires organizations to be agile and visionary in this arena to grow with the ever-changing technological, regulatory, and unique customer requirements, while simultaneously ensuring data security and privacy. To that end, the SDSC Health CI Division is participating in a multi-institution project with City of Hope, a National Cancer Institute (NCI)-designated Comprehensive Cancer Center, and other universities and health organizations to create a research cyberinfrastructure that includes a secure, cloud-based data management platform. The data management platform consolidates all CTS data, including datasets, codes, and accompanying documentation, and operates within Sherlock Cloud, a secure hybrid cloud platform. The innovative platform allows every member of the CTS team to securely access and use all CTS data and information in real time in a consolidated, integrated, and secure manner. Researchers have access to a data warehouse, domain-specific data marts, and analytics platform built on secure cloud technology that greatly enhances the reliability and accuracy of the data collected, and has a more seamless mechanism to access, annotate, input, and transmit data, thereby heightening accuracy and the quality of collaborative analysis performed. This approach not only transforms how CTS data is collected, stored, and shared for high-impact research, but also has the potential to reduce associated costs of ongoing research while increasing efficiency and security. SDSC Health CI Division’s work in building innovative data management and compute solutions for health care researchers over the last decade, and its collaboration with the CTS researchers more recently, has demonstrated a clear need for a unique type of managed services capability that serves the biomedical community. Specifically, the epidemiology cohorts have practiced a more traditional approach consisting of decentralized data management and computing. Even though this approach has served the community well in the past, as we look into the future, it has obvious shortcomings in terms of supporting enterprise-level data management, scale, interoperability, and provenance. SDSC Health CI Division takes a modern approach to information lifecycle management through its Sherlock Cloud platform. Sherlock Cloud Data Management framework captures users’ evolving requirements, and provides centralized, secure cloud-based managed services capability that supports end-to-end data lifecycle management including data integration capabilities, allowing data to be captured, rationalized, homogenized, and managed using best practices and standards. This includes dynamic mappings, transformations, and master data-management techniques utilizing established governance methodologies, with specific focus on data quality, metadata management, and governance. The CTS data management and analytics platform implementation can serve as a model for other epidemiology cohorts and studies. SDSC Health CI Division plans to achieve this by leveraging the existing investment NCI and CTS has already made in building the cloud-based, secure CTS data management platform. Using container technology, Sherlock will package the various components of the data management framework and analytical tools, creating a turnkey offering that can be deployed for other similar studies. This cloud platform agnostic, open-source, turnkey solution provides other cohorts a template they can build on top of, and leverage all the work CTS has performed to develop the core, domain-specific data management and analytics capability, and all that will remain is study specific customizations that cohorts will need to perform. Additionally, APIs built on top of the in-built data model will provide seamless integration with external NCI- and NIH-established Commons. The epidemiology community has a way to go before it is on par with communities in other verticals in their adoption of cutting-edge, enterprise platforms and tools for data management and analytics, but initiatives like CTS can demonstrate that, once designed and developed, these capabilities can be easily packaged and deployed for other studies, providing a framework that a larger research community can leverage. Citation Format: Sandeep Chandra. California Teachers Study (CTS) Data Management Platform: A model for a repeatable turnkey, end-to-end, cloud-based data management and analytics solution for epidemiology cohorts [abstract]. In: Proceedings of the AACR Special Conference on Modernizing Population Sciences in the Digital Age; 2019 Feb 19-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2020;29(9 Suppl):Abstract nr IA20.

CS-CAVE

Bits & Bytes CS-CAVE: DISTRICTS’ ROLES IN BROADENING CS ACCESS Susan Yonezawa, Nan Renner, Monica Sweet, Beth Simon and Diane Baxter This research investigates the spread and sustainability of university efforts to increase the number and types of computer science (CS) courses available in K-12. Two university/K-12 collaborative projects in the San Diego region provide insights into pathways for sustainability. Nearly 75 teachers were trained to teach CS Principles through professional development by university CS faculty and staff. Despite myriad challenges, one-third of the teachers implemented the course. Subsequent research explores three school districts as active players rather than passive recipients (or worse, resistors) of NSF-funded university-led efforts, revealing political and cultural challenges to implementation and sustainability of CS education reform. Computing pervades contemporary life. The US economy demands workers with computational competence. Yet, few K-12 students, particularly those from low-income and minority back- grounds, have access to formal learning opportunities in computer science (CS) [2]. To meet economic demands and give students access to opportunities, the US education system must expand CS offerings in K-12 education. Equally critical, this expansion must be equitable and accessible to diverse student populations, not limited just to the affluent and tech-endowed schools. Educational leaders call for expanding access to K-12 comput- ing education, with deliberate attention to broadening participation by students traditionally underrepresented in computer science, including female, African American, Latino, Native American, and students with disabilities/learning differences. Researchers have documented disparities in students’ K-12 access to computer science learning opportunities, but higher education’s investment to actively assist K-12 is more recent [1]. Higher education part- nerships with K-12 have increased the number and types of CS courses available for students. Our research expands the scope to investigate sustainability and spread of CS reforms. Two sequential CS education projects involving collaboration between California higher education (UC San Diego and San 80 acm Inroads 2015 December • Vol. 6 • No. 4 Working with district partners, ComPASS and CS-CaVE reveal that the technical as- pects of CS curriculum and teacher training present only one set of challenges. Diego State University) and K-12 education in the San Diego region highlight this recent progression from implementation to sustainability research in K-12 CS education, providing insight into pathways for sustainability. Two efforts to broaden participation in CS: From ComPASS to CS-CaVE In 2011, the National Science Foundation (NSF) funded teams at University of California San Diego (UCSD) and San Diego State University (SDSU) to conduct research on adapting a college- level course, CS Principles (CSP), for high schools in San Diego County, California. This CE21 project, “Computing Principles for All Students’ Success” (ComPASS), aimed to create a broader pool of CSP teachers, expand CSP class offerings, increase the number of schools offering CSP, and broaden teacher and student partici- pation. ComPASS project objectives were to: ■ develop and evaluate content and pedagogy training and support resources for teachers (with or without computing backgrounds) to teach CSP; ■ build a regional computing education community to provide sustainable peer support for new CSP teachers. ComPASS has trained approximately 75 teachers to teach CSP, through intensive CS workshops and professional development led by UCSD and SDSU Computer Science faculty and instructors from the San Diego Supercomputer Center (SDSC). However,

Seamless Synthetic Aperture Radar Archive for Interferometry Analysis

Abstract. The NASA Advancing Collaborative Connections for Earth System Science (ACCESS) seamless synthetic aperture radar (SAR) archive (SSARA) project is a collaboration between UNAVCO, the Alaska Satellite Facility (ASF), the Jet Propulsion Laboratory (JPL), and OpenTopography at the San Diego Supercomputer Center (SDSC) to design and implement a seamless distributed access system for SAR data and derived interferometric SAR (InSAR) data products. A unified application programming interface (API) has been created to search the SAR archives at ASF and UNAVCO, 30 and 90-m SRTM DEM data available through OpenTopography, and tropospheric data from the NASA OSCAR project at JPL. The federated query service provides users a single access point to search for SAR granules, InSAR pairs, and corresponding DEM and tropospheric data products from the four archives, as well as the ability to search and download pre-processed InSAR products from ASF and UNAVCO.

Implementations of Urgent Computing on Production HPC Systems

The requirements of urgent computing applications distinguish those applications from other jobs running within a production batch system. To accomodate urgent computing, specific strategies may be used to allow fast job start time, while mitigating adverse effects on the other parts of the workload. San Diego Supercomputer Center (SDSC) has provided high-performance computing resources in a production-oriented environment to urgent computing appli-cations using several techniques. Processing of data, triggered by real-world disaster events has been accomplished. An on-demand reservation capability to support urgent computing applications has been developed and is currently available on the SDSC Trestles machine.

Chronopolis Digital Preservation Network

The Chronopolis Digital Preservation Initiative, one of the Library of Congress’ latest efforts to collect and preserve at-risk digital information, has completed its first year of service as a multi-member partnership to meet the archival needs of a wide range of domains.Chronopolis is a digital preservation data grid framework developed by the San Diego Supercomputer Center (SDSC) at UC San Diego, the UC San Diego Libraries (UCSDL), and their partners at the National Center for Atmospheric Research (NCAR) in Colorado and the University of Maryland's Institute for Advanced Computer Studies (UMIACS).Chronopolis addresses a critical problem by providing a comprehensive model for the cyberinfrastructure of collection management, in which preserved intellectual capital is easily accessible, and research results, education material, and new knowledge can be incorporated smoothly over the long term. Integrating digital library, data grid, and persistent archive technologies, Chronopolis has created trusted environments that span academic institutions and research projects, with the goal of long-term digital preservation.A key goal of the Chronopolis project is to provide cross-domain collection sharing for long-term preservation. Using existing high-speed educational and research networks and mass-scale storage infrastructure investments, the partnership is leveraging the data storage capabilities at SDSC, NCAR, and UMIACS to provide a preservation data grid that emphasizes heterogeneous and highly redundant data storage systems.In this paper we will explore the major themes within Chronopolis, including:a) The philosophy and theory behind a nationally federated data grid for preservation. b) The core tools and technologies used in Chronopolis. c) The metadata schema that is being developed within Chronopolis for all of the data elements. d) Lessons learned from the first year of the project.e) Next steps in digital preservation using Chronopolis: how we plan to strengthen and broaden our network with enhanced services and new customers.

NDIIPP Models for Mass Data Transmission and Storage

Preservation of digital content into the future will rely on the ability of institutions to provide robust system infrastructures that leverage the use of distributed and shared services and tools. The academic, nonprofit, and government entities that make up the National Digital Information Infrastructure and Preservation Program (NDIIPP) partner network have been working toward an architecture that can provide for reliable redundant geographically dispersed copies of their digital content. The NDIIPP program has conducted a set of initiatives that have enabled partners to better understand the requirements for effective collection interchange. The NDIIPP program partnered with the San Diego Supercomputer Center (SDSC) to determine the feasibility of data transmission and storage utilizing the best of breed technologies inherent to U.S. high-speed research networks and high-performance computing data storage infrastructures. The results of this partnership guided the development of the Library of Congress's cyberinfrastructure and its approach to network data transfer. Other NDIIPP partners, too, are researching a range of network architecture models for data exchange and storage. All of these explorations will build toward the development of best practices for sustainable interoperability and storage solutions.

Digital Preservation Theory and Application: Transcontinental Persistent Archives Testbed Activity

The National Archives and Records Administration (NARA) and EU SHAMAN projects are working with multiple research institutions on tools and technologies that will supply a comprehensive, systematic, and dynamic means for preserving virtually any type of electronic record, free from dependence on any specific hardware or software. This paper describes the joint development work between the University of Liverpool and the San Diego Supercomputer Center (SDSC) at the University of California, San Diego on the NARA and SHAMAN prototypes. The aim is to provide technologies in support of the required generic data management infrastructure. We describe a Theory of Preservation that quantifies how communication can be accomplished when future technologies are different from those available at present. This includes not only different hardware and software, but also different standards for encoding information. We describe the concept of a “digital ontology” to characterize preservation processes; this is an advance on the current OAIS Reference Model of providing representation information about records. To realize a comprehensive Theory of Preservation, we describe the ongoing integration of distributed shared collection management technologies, digital library browsing, and presentation technologies for the NARA and SHAMAN Persistent Archive Testbeds.

Identification and Role of the Homodimerization Interface of the Glycosylphosphatidylinositol-anchored Membrane Type 6 Matrix Metalloproteinase (MMP25)

The membrane type (MT) 6 matrix metalloproteinase (MMP) (MMP25) is a glycosylphosphatidylinositol-anchored matrix metalloproteinase (MMP) that is highly expressed in leukocytes and in some cancer tissues. We previously showed that natural MT6-MMP is expressed on the cell surface as a major reduction-sensitive form of M(r) 120, likely representing enzyme homodimers held by disulfide bridges. Among the membrane type-MMPs, the stem region of MT6-MMP contains three cysteine residues at positions 530, 532, and 534 which may contribute to dimerization. A systematic site-directed mutagenesis study of the Cys residues in the stem region shows that Cys(532) is involved in MT6-MMP dimerization by forming an intermolecular disulfide bond. The mutagenesis data also suggest that Cys(530) and Cys(534) form an intramolecular disulfide bond. The experimental observations on cysteines were also investigated by computational studies of the stem peptide, which validate these proposals. Dimerization is not essential for transport of MT6-MMP to the cell surface, partitioning into lipid rafts or cleavage of alpha-1-proteinase inhibitor. However, monomeric forms of MT6-MMP exhibited enhanced autolysis and metalloprotease-dependent degradation. Collectively, these studies establish the stem region of MT6-MMP as the dimerization interface, an event whose outcome imparts protease stability to the protein.

CARMEN: Distributed Search and Data-Mining in Complex Neurophysiology Data

Event Abstract Back to Event CARMEN: Distributed Search and Data-Mining in Complex Neurophysiology Data Bojian Liang1*, Tom Jackson1, Martyn Fletcher1, Mark Jessop1 and Jim Austin1 1 University of York, United Kingdom We describe research within the UK CARMEN [1] e-Science project which is addressing the challenge of creating and managing experimental data and methods within the context of Neuroscience research. The traditional research approach of testing a hypothesis and publishing the results is hampered in situations where others need to build on the results and need access to the data or original methods. CARMEN addresses this issue by allowing scientists to share data & methods within a collaborative Grid environment. The CARMEN platform (a CAIRN) is a Grid-based, shared data and services repository. Central to the data management challenge is providing visualisation and search capabilities across the stored, high-volume, time-series, spike train data from nerve cell recordings. A prior UK e-Science project, DAME[2], developed a distributed signal search engine, called Signal Data Explorer (SDE) [3]. Interoperability between SDE and other neuroscience software services is being investigated within CARMEN, to provide a centralised capability for visualisation and search. The SDE provides data visualisation, transformations and real-time search capabilities for complex time-series signal data. The interactive search capability is provided by the SDE via distributed Grid data repositories which use: • Distributed data management - SDE interfaces directly to the Storage Resource Broker (SRB) from San Diego Supercomputer Center (SDSC); • The pattern matching architecture consisting of Pattern Match Controller (PMC) and Pattern Match Engine (PME) components. The raw data to be stored by the CARMEN CAIRN nodes consists of spatiotemporal signals expressed, either as time-series recordings (collected by single electrodes or MEAs), or as image files. Querying of the data at the “signal” level is provided via a highly intuitive and interactive SDE GUI. This GUI allows users to formulate queries by simple selection of regions of interest in time-series signals, or via libraries of templates queries, built up over the course of prior experiments. The SDE GUI also supports complex visualisation requirements, permitting users to explore data events and features in detail or in a broad overview, as required. Auxiliary viewing windows permit users to arrange collated views on data signals, and support additional functionality such as signal superimposition, filtering and spike sorting. The talk will provide a demonstration of progress on the development of the search and visualisation capabilities, working on experimental MEA data, and will show how additional analysis functionality, such as spike sorting or correlation functions, can be supported within the tool. User feedback is an important aspect of the development iteration and we will conclude with examples of how SDE is supporting laboratory experiments within the CARMEN consortium.

Mining performance data for metascheduling decision support in the Grid

Metaschedulers in the Grid need dynamic information to support their scheduling decisions. Job response time on computing resources, for instance, is such a performance metric. In this paper, we propose an Instance Based Learning technique to predict response times by mining historical performance data. The novelty of our approach is to introduce policy attributes in representing and comparing resource states, which are defined as the pools of running and queued jobs on the resources at the time of making predictions. The policy attributes reflect the local scheduling policies and they can be automatically discovered using genetic search. An extensive empirical evaluation is conducted to validate our technique using real workload traces, which are collected from the NIKHEF production cluster on the LHC Computing Grid and Blue Horizon in the San Diego Supercomputer Center (SDSC). The experimental results show that acceptable prediction accuracy can be achieved, where the normalized average prediction errors for response times are ranging from 0.57 to 0.79.

The SCEC Community Modeling Environment: An Information Infrastructure for System-level Earthquake Science

In the July/August issue of SRL , Tom Owens, writing his first column for the Electronic Seismologist (ES), asked “The Question”: Can we as a community develop the long-term relationships with the computer science community, the financial resources, and the culture necessary to build and maintain a state-of-the-art information technology (IT) structure that will really make a difference in the way we do our science? He discussed the tough practical challenges posed by The Question and highlighted some of the IT efforts underway by IRIS, EarthScope, and other geoscience organizations. In particular, he mentioned a new IT project by the Southern California Earthquake Center (SCEC) that involves an interesting mix of geoscientists and computer scientists. As a follow-up, he invited us to contribute an article to ES on how this project was formulated, how the collaborations have been set up, and how it's going. Tall order, but we'll try! We begin with a little background. In October 2001, SCEC was awarded a five-year, $10 million grant by the NSF Information Technology Research (ITR) Program to develop a Community Modeling Environment (CME)—what IT aficionados would call a “collaboratory”—for system-level earthquake science. In addition to a number of participating SCEC institutions, the project involves IRIS (http://www.iris.edu), the USGS' Pasadena office (http://pasadena.wr.usgs.gov), and two major IT research organizations, the Information Sciences Institute (ISI) of the University of Southern California (http://www.isi.edu) and the San Diego Supercomputer Center (SDSC) of the University of California at San Diego (http://www.sdsc.edu). The overarching goal of the CME project is to improve the information infrastructure for seismic hazard analysis (SHA), which has historically been the major focus of SCEC's research program. SHA seeks to describe the maximum level of shaking that can be expected at a specified site on the Earth's surface due to earthquakes anticipated over a …

Long-term data preservation efforts ramping up

The paper discusses long-term archiving, preservation and access to digital data. From the US National Archives, and the San Diego Supercomputer Center (SDSC) to the national libraries of the Netherlands and Australia, archivists, computer scientists, and librarians are pioneering the technologies that might someday become the methods our descendants use to access every sort of record of our time.

Using Apples to Schedule Simple SARA on the Computational Grid

Computational Grids, composed of distributed and often heterogeneous computing resources, have become the platform of choice for many performance-challenged applications. Proof-of-concept implementations have demonstrated that both Grids and clustered environments have the potential to provide great performance benefits to distributed resource-intensive applications. However, at the present time, careful staging, scheduling, and/or reservation of resources is essential in order for applications to achieve performance in Grid environments. If Computational Grids and shared computational clusters are to achieve their full potential, it must be possible for users to achieve application performance at any given time, and when other users are present in the system. In this paper, we describe the initial development of an AppLeS (application-level scheduler) for the resource selection portion of the Synthetic Aperture Radar Atlas (SARA) application, developed at the Jet Propulsion Laboratory (JPL) and the San Diego Supercomputer Center (SDSC). We demonstrate the effectiveness of application scheduling for distributed data applications such as SARA by providing a performance-efficient strategy for retrieving SARA data files in everyday, multiple-user Grid environments.