Achieving Maintainable Cross-Platform Performance in the Particle-in-Cell Accelerator Modeling Code Synergia using Kokkos
Explore the source record for details and available documents.
Engineering topics
Publications and source records attributed to Lu, Qiming.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
BigData Express: a schedulable, predictable, and high-performance data transfer service. A peer-to-peer, scalable, and extensible data transfer model; A visually appealing, easy-to-use web portal; A high-performance data transfer engine; On-demand provisioning of end-to-end network paths with guaranteed QoS; Robust and flexible error handling; CILogon-based security (certificate version); A rich set of REST APIs to support scientific workflows
Increasingly, scientific research discoverydepends on the analysis of extremely large amounts of data. Thisdata is gathered through very large scientific instruments, verylarge numbers of small devices, and mid-scale instruments.These devices generate exponentially increasing volumes of data.Such data is often obtained, analyzed, stored, visualized, andtransferred at and among sites world-wide by largecollaborations of science communities. Managing and movingextremely large volumes of data across today’s networks is aspecial multidimensional challenge. Also, existing datamanagement and movement services and tools often areinadequate to address all specific requirements. This paperdescribes the various components of this challenge and howthose issues can be addressed by ROBIN, a uniquecomprehensive set of integrated services designed specificallyfor managing and moving extremely large amounts of data overlong distances, e.g., thousands of miles around the globe. Thispaper also describes the results of initial experiments using thatset of integrated services.
HEPCloud is rapidly becoming the primary system for provisioning compute resources for all Fermilab-affiliated experiments. In order to reliably meet the peak demands of the next generation of High Energy Physics experiments, Fermilab must plan to elastically expand its computational capabilities to cover the forecasted need. Commercial cloud and allocation-based High Performance Computing (HPC) resources both have explicit and implicit costs that must be considered when deciding when to provision these resources, and at which scale. In order to support such provisioning in a manner consistent with organizational business rules and budget constraints, we have developed a modular intelligent decision support system (IDSS) to aid in the automatic provisioning of resources spanning multiple cloud providers, multiple HPC centers, and grid computing federations. In this paper, we discuss the goals and architecture of the HEPCloud Facility, the architecture of the IDSS, and our early experience in using the IDSS for automated facility expansion both at Fermi and Brookhaven National Laboratory.
Big Data has emerged as a driving force for scientific discoveries. Large scientific instruments (e.g., colliders, and telescopes) generate exponentially increasing volumes of data. To enable scientific discovery, science data must be collected, indexed, archived, shared, and analyzed, typically in a widely distributed, highly collaborative manner. Data transfer is now an essential function for science discoveries, particularly within big data environments. Although significant improvements have been made in the area of bulk data transfer, the currently available data transfer tools and services can not successfully address the high-performance and time-constraint challenges of data transfer required by extreme-scale science applications for the following reasons: disjoint end-to-end data transfer loops, cross-interference between data transfers, and existing data transfer tools and services are oblivious to user requirements (deadline and QoS requirements). Fermilab has been working on the BigData Express project to address these problems. BigData Express seeks to provide a schedulable, predictable, and high-performance data transfer service for big data science. The BigData Express software is being deployed and evaluated at multiple research institutions, which include UMD, StarLight, FNAL, KISTI, KSTAR, SURFnet, Ciena, and other sites. Meanwhile, the BigData Express research team is collaborating with the StarLight International/National Communications Exchange Facility to deploy BigData Express at various research platforms, including Pacific Research Platform, National Research Platform, and Global Research Platform. It is envisioned that we are working toward building a high-performance data transfer federation for big data science.
The Decision Engine is a critical component of the HEP Cloud Facility. It provides the functionality of resource scheduling for disparate resource providers, including those which may have a cost or a restricted allocation of cycles. Along with the architecture, design, and requirements for the Decision Engine, this document will provide the rationale and explanations for various design decisions. In some cases, requirements and interfaces for a limited subset of external services will be included in this document. This document is intended to be a high level design. The design represented in this document is not complete and does not break everything down in detail. The class structures and pseudo-code exist for example purposes to illustrate desired behaviors, and as such, should not be taken literally. The protocols and behaviors are the important items to take from this document. This project is still in prototyping mode so flaws and inconsistencies may exist and should be noted and treated as failures.
The study of the response of complex dynamical social, biological, or technological networks to external perturbations has numerous applications. Random Boolean Networks (RBNs) are commonly used a simple generic model for certain dynamics of complex systems. Traditionally, RBNs are interconnected randomly and without considering any spatial extension and arrangement of the links and nodes. However, most real-world networks are spatially extended and arranged with regular, power-law, small-world, or other non-random connections. Here we explore the RBN network topology between extreme local connections, random small-world, and pure random networks, and study the damage spreading with small perturbations. We find that spatially local connections change the scaling of the relevant component at very low connectivities ($\bar{K} \ll 1$) and that the critical connectivity of stability $K_s$ changes compared to random networks. At higher $\bar{K}$, this scaling remains unchanged. We also show that the relevant component of spatially local networks scales with a power-law as the system size N increases, but with a different exponent for local and small-world networks. The scaling behaviors are obtained by finite-size scaling. We further investigate the wiring cost of the networks. From an engineering perspective, our new findings provide the key design trade-offs between damage spreading (robustness), the network's wiring cost, and the network's communication characteristics.
Abstract Not Provided
We consider multiphysics applications from algorithmic and architectural perspectives, where ‘‘algorithmic’’ includes both mathematical analysis and computational complexity, and ‘‘architectural’’ includes both software and hardware environments. Many diverse multiphysics applications can be reduced, en route to their computational simulation, to a common algebraic coupling paradigm. Mathematical analysis of multiphysics coupling in this form is not always practical for realistic applications, but model problems representative of applications discussed herein can provide insight. A variety of software frameworks for multiphysics applications have been constructed and refined within disciplinary communities and executed on leading-edge computer systems. We examine several of these, expose some commonalities among them, and attempt to extrapolate best practices to future systems. From our study, we summarize challenges and forecast opportunities.
A complex running system, such as the NOvA online data acquisition, consists of a large number of distributed but closely interacting components. This paper describes a generic real-time correlation analysis and event identification engine, named Message Analyzer. Its purpose is to capture run time abnormalities and recognize system failures based on log messages from participating components. The initial design of analysis engine is driven by the data acquisition (DAQ) of the NOvA experiment. The Message Analyzer performs filtering and pattern recognition on the log messages and reacts to system failures identified by associated triggering rules. The tool helps the system maintain a healthy running state and to minimize data corruption. This paper also describes a domain specific language that allows the recognition patterns and correlation rules to be specified in a clear and flexible way. In addition, the engine provides a plugin mechanism for users to implement specialized patterns or rules in generic languages such as C++.
The message logging system provides the infrastructure for all of the distributed processes in the data acquisition (DAQ) to report status messages of various severities in a consistent manner to a central location, as well as providing the tools for displaying and archiving the messages. The message logging system has been developed over a decade, and has been run successfully on CDF and CMS experiments. The most recent work to the message logging system is to build it as a stand-alone package with the name MessageFacility which works for any generic framework or applications, with NO A as the first driving user. System designs and architectures, as well as the efforts of making it a generic library will be discussed. We also present new features that have been added.
Random Boolean Networks (RBNs) are often used as generic models for certain dynamics of complex systems, ranging from social networks, neural networks, to gene or protein interaction networks. Traditionally, RBNs are interconnected randomly and without considering any spatial arrangement of the links and nodes. However, most real-world networks are spatially extended and arranged with regular, small-world, or other non-random connections. Here we explore the RBN network topology between extreme local connections, random small-world, and random networks, and study the damage spreading with small perturbations. We find that spatially local connections change the scaling of the relevant component at very low connectivities ({bar K} << 1) and that the critical connectivity of stability K{sub s} changes compared to random networks. At higher {bar K}, this scaling remains unchanged. We also show that the relevant component of spatially local networks scales with a power-law as the system size N increases, but with a different exponent for local and small-world networks. The scaling behaviors are obtained by finite-size scaling. We further investigate the wiring cost of the networks. From an engineering perspective, our new findings provide the key trade-offs between damage spreading (robustness), the network wiring cost, and the network's communication characteristics.