Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “workload”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 991 records · Page 55

Managing variability in the IO performance of petascale storage systems.

Significant challenges exist for achieving peak or even consistent levels of performance when using IO systems at scale. They stem from sharing IO system resources across the processes of single large-scale applications and/or multiple simultaneous programs causing internal and external interference, which in turn, causes substantial reductions in IO performance. This paper presents interference effects measurements for two different file systems at multiple supercomputing sites. These measurements motivate developing a 'managed' IO approach using adaptive algorithms varying the IO system workload based on current levels and use areas. An implementation of these methods deployed for the shared, general scratch storage system on Oak Ridge National Laboratory machines achieves higher overall performance and less variability in both a typical usage environment and with artificially introduced levels of 'noise'. The latter serving to clearly delineate and illustrate potential problems arising from shared system usage and the advantages derived from actively managing it.

99 GENERAL AND MISCELLANEOUS↗

Active Flash: Performance-Energy Tradeoffs for Out-of-Core Processing on Non-Volatile Memory Devices

In this abstract, we study the performance and energy tradeoffs involved in migrating data analysis into the flash device, a process we refer to as Active Flash. The Active Flash paradigm is similar to 'active disks', which has received considerable attention. Active Flash allows us to move processing closer to data, thereby minimizing data movement costs and reducing power consumption. It enables true out-of-core computation. The conventional definition of out-of-core solvers refers to an approach to process data that is too large to fit in the main memory and, consequently, requires access to disk. However, in Active Flash, processing outside the host CPU literally frees the core and achieves real 'out-of-core' analysis. Moving analysis to data has long been desirable, not just at this level, but at all levels of the system hierarchy. However, this requires a detailed study on the tradeoffs involved in achieving analysis turnaround under an acceptable energy envelope. To this end, we first need to evaluate if there is enough computing power on the flash device to warrant such an exploration. Flash processors require decent computing power to run the internal logic pertaining to the Flash Translation Layer (FTL), which is responsible for operations such as address translation, garbage collection (GC) and wear-leveling. Modern SSDs are composed of multiple packages and several flash chips within a package. The packages are connected using multiple I/O channels to offer high I/O bandwidth. SSD computing power is also expected to be high enough to exploit such inherent internal parallelism within the drive to increase the bandwidth and to handle fast I/O requests. More recently, SSD devices are being equipped with powerful processing units and are even embedded with multicore CPUs (e.g. ARM Cortex-A9 embedded processor is advertised to reach 2GHz frequency and deliver 5000 DMIPS; OCZ RevoDrive X2 SSD has 4 SandForce controllers, each with 780MHz max frequency Tensilica core). Efforts that take advantage of the available computing cycles on the processors on SSDs to run auxiliary tasks other than actual I/O requests are beginning to emerge. Kim et al. investigate database scan operations in the context of processing on the SSDs, and propose dedicated hardware logic to speed up scans. Also, cluster architectures have been explored, which consist of low-power embedded CPUs coupled with small local flash to achieve fast, parallel access to data. Processor utilization on SSD is highly dependent on workloads and, therefore, they can be idle during periods with no I/O accesses. We propose to use the available processing capability on the SSD to run tasks that can be offloaded from the host. This paper makes the following contributions: (1) We have investigated Active Flash and its potential to optimize the total energy cost, including power consumption on the host and the flash device; (2) We have developed analytical models to analyze the performance-energy tradeoffs for Active Flash, by treating the SSD as a blackbox, this is particularly valuable due to the proprietary nature of the SSD internal hardware; and (3) We have enhanced a well-known SSD simulator (from MSR) to implement 'on-the-fly' data compression using Active Flash. Our results provide a window into striking a balance between energy consumption and application performance.

42 ENGINEERING↗

Building a Parallel Cloud Storage System using OpenStack’s Swift Object Store and Transformative Parallel I/O

Our project consists of bleeding-edge research into replacing the traditional storage archives with a parallel, cloud-based storage solution. It used OpenStack's Swift Object Store cloud software. It's Benchmarked Swift for write speed and scalability. Our project is unique because Swift is typically used for reads and we are mostly concerned with write speeds. Cloud Storage is a viable archive solution because: (1) Container management for larger parallel archives might ease the migration workload; (2) Many tools that are written for cloud storage could be utilized for local archive; and (3) Current large cloud storage practices in industry could be utilized to manage a scalable archive solution.

99 GENERAL AND MISCELLANEOUS↗

QoS support for end users of I/O-intensive applications using shared storage systems

I/O-intensive applications are becoming increasingly common on today's high-performance computing systems. While performance of compute-bound applications can be effectively guaranteed with techniques such as space sharing or QoS-aware process scheduling, it remains a challenge to meet QoS requirements for end users of I/O-intensive applications using shared storage systems because it is difficult to differentiate I/O services for different applications with individual quality requirements. Furthermore, it is difficult for end users to accurately specify performance goals to the storage system using I/O-related metrics such as request latency or throughput. As access patterns, request rates, and the system workload change in time, a fixed I/O performance goal, such as bounds on throughput or latency, can be expensive to achieve and may not lead to a meaningful performance guarantees such as bounded program execution time. We propose a scheme supporting end-users QoS goals, specified in terms of program execution time, in shared storage environments. We automatically translate the users performance goals into instantaneous I/O throughput bounds using a machine learning technique, and use dynamically determined service time windows to efficiently meet the throughput bounds. We have implemented this scheme in the PVFS2 parallel file system and have conducted an extensive evaluation. Our results show that this scheme can satisfy realistic end-user QoS requirements by making highly efficient use of the I/O resources. The scheme seeks to balance programs attainment of QoS requirements, and saves as much of the remaining I/O capacity as possible for best-effort programs.

97 MATHEMATICS AND COMPUTING↗

Petascale system management experiences.

Petascale High-Performance Computing (HPC) systems are among the largest systems in the world. Intrepid, one such system, is a 40,000 node, 556 teraflop Blue Gene/P system that has been deployed at Argonne National Laboratory. In this paper, we provide some background about the system and our administration experiences. In particular, due to the scale of the system, we have faced a variety of issues, some surprising to us, that are not common in the commodity world. We discuss our expectations, these issues, and approaches we have used to address them. HPC systems are a bellwether for computing systems at large, in multiple regards. HPC users are motivated by the need for absolute performance; this results in two important pushes. HPC users are frequently early adopters of new technologies and techniques. Successful technologies, like Infiniband, prove their value in HPC before gaining wider adoption. Unfortunately, this early adoption alone is not sufficient to achieve the levels of performance required by HPC users; parallelism must also be harnessed. Over the last 15 years, beowulf clustering has provided amazing accessibility to non-HPC-savvy and even non-technical audiences. During this time, substantial adoption of clustering has occurred in many market segments unrelated to computational science. A simple trend has emerged: the scale and performance of high-end HPC systems are uncommon at first, but become commonplace over the course of 3-5 years. For example, in early 2003, several systems on the Top500 list consisted of either 1024 nodes or 4096-8192 cores. In 2008, such systems are commonplace. The most recent generation of high-end HPC systems, so called petascale systems, are the culmination of years of research and development in research and academia. Three such systems have been deployed thus far. In addition to the 556 TF Intrepid system at Argonne National Laboratory, a 596 TF Blue Gene/L-based system has been deployed at Lawrence Livermore National Laboratory, and a 504 TF Opteron-based system has been deployed at Texas Advanced Computing Center (TACC). Intrepid is comprised of 40,960 nodes with a total of 163,840 cores. While systems like these are uncommon now, we expect them to become more widespread in the coming years. The scale of these large systems impose several requirements upon system architecture. The need for scalability is obvious, however, power efficiency and density constraints have become increasingly important in recent years. At the same time, because the size of administrative staff cannot grow linearly with the system size, more efficient system management techniques are needed. In this paper we will describe our experiences administering Intrepid. Over the last year, we have experienced a number of interesting challenges in this endeavor. Our initial expectation was for scalability to be the dominant system issue. This expectation was not accurate. Several issues expected to have minor impact have played a much greater role in system operations. Debugging, due to the large numbers of components used in scalable system operations, has become a much more difficult endeavor. The system has a sophisticated monitoring system, however, the analysis of this data has been problematic. These issues are not specific to HPC workloads in any way, so we expect them to be of general interest. This paper consists of three major parts. First, we will provide a detailed overview of several important aspects of Intrepid's hardware and software. In this, we will highlight aspects that have featured prominently in our system management experiences. Next, we will describe our administration experiences in detail. Finally, we will draw some conclusions based on these experiences. In particular, we will discuss the implications for the non-HPC world, system managers, and system software developers.

99 GENERAL AND MISCELLANEOUS↗

Parallel In Situ Indexing for Data-intensive Computing

As computing power increases exponentially, vast amount of data is created by many scientific re- search activities. However, the bandwidth for storing the data to disks and reading the data from disks has been improving at a much slower pace. These two trends produce an ever-widening data access gap. Our work brings together two distinct technologies to address this data access issue: indexing and in situ processing. From decades of database research literature, we know that indexing is an effective way to address the data access issue, particularly for accessing relatively small fraction of data records. As data sets increase in sizes, more and more analysts need to use selective data access, which makes indexing an even more important for improving data access. The challenge is that most implementations of in- dexing technology are embedded in large database management systems (DBMS), but most scientific datasets are not managed by any DBMS. In this work, we choose to include indexes with the scientific data instead of requiring the data to be loaded into a DBMS. We use compressed bitmap indexes from the FastBit software which are known to be highly effective for query-intensive workloads common to scientific data analysis. To use the indexes, we need to build them first. The index building procedure needs to access the whole data set and may also require a significant amount of compute time. In this work, we adapt the in situ processing technology to generate the indexes, thus removing the need of read- ing data from disks and to build indexes in parallel. The in situ data processing system used is ADIOS, a middleware for high-performance I/O. Our experimental results show that the indexes can improve the data access time up to 200 times depending on the fraction of data selected, and using in situ data processing system can effectively reduce the time needed to create the indexes, up to 10 times with our in situ technique when using identical parallel settings.

97 MATHEMATICS AND COMPUTING↗

Adaptive real-time methodology for optimizing energy-efficient computing

Dynamic voltage and frequency scaling (DVFS) is an effective way to reduce energy and power consumption in microprocessor units. Current implementations of DVFS suffer from inaccurate modeling of power requirements and usage, and from inaccurate characterization of the relationships between the applicable variables. A system and method is proposed that adjusts CPU frequency and voltage based on run-time calculations of the workload processing time, as well as a calculation of performance sensitivity with respect to CPU frequency. The system and method are processor independent, and can be applied to either an entire system as a unit, or individually to each process running on a system.

97 MATHEMATICS AND COMPUTING↗

Production Experiences with the Cray-Enabled TORQUE Resource Manager

High performance computing resources utilize batch systems to manage the user workload. Cray systems are uniquely different from typical clusters due to Cray s Application Level Placement Scheduler (ALPS). ALPS manages binary transfer, job launch and monitoring, and error handling. Batch systems require special support to integrate with ALPS using an XML protocol called BASIL. Previous versions of Adaptive Computing s TORQUE and Moab batch suite integrated with ALPS from within Moab, using PERL scripts to interface with BASIL. This would occasionally lead to problems when all the components would become unsynchronized. Version 4.1 of the TORQUE Resource Manager introduced new features that allow it to directly integrate with ALPS using BASIL. This paper describes production experiences at Oak Ridge National Laboratory using the new TORQUE software versions, as well as ongoing and future work to improve TORQUE.

Adaptive Comput- ing↗

FlexiWay: A Cache Energy Saving Technique Using Fine-grained Cache Reconfiguration

Recent trends of CMOS scaling and use of large last level caches (LLCs) have led to significant increase in the leakage energy consumption of LLCs and hence, managing their energy consumption has become extremely important in modern processor design. The conventional cache energy saving techniques require offline profiling or provide only coarse granularity of cache allocation. We present FlexiWay, a cache energy saving technique which uses dynamic cache reconfiguration. FlexiWay logically divides the cache sets into multiple (e.g. 16) modules and dynamically turns off suitable and possibly different number of cache ways in each module. FlexiWay has very small implementation overhead and it provides fine-grain cache allocation even with caches of typical associativity, e.g. an 8-way cache. Microarchitectural simulations have been performed using an x86-64 simulator and workloads from SPEC2006 suite. Also, FlexiWay has been compared with two conventional energy saving techniques. The results show that FlexiWay provides largest energy saving and incurs only small loss in performance. For single, dual and quad core systems, the average energy saving using FlexiWay are 26.2%, 25.7% and 22.4%, respectively.

Cache leakage energy saving↗

Layout-Aware I/O Scheduling for Terabits Data Movement

Many science facilities, such as the Department of Energy s Leadership Computing Facilities and experimental facilities including the Spallation Neutron Source, Stanford Linear Accelerator Center, and Advanced Photon Source, produce massive amounts of experimental and simulation data. These data are often shared among the facilities and with collaborating institutions. Moving large datasets over the wide- area network (WAN) is a major problem inhibiting collaboration. Next- generation, terabit-networks will help alleviate the problem, however, the parallel storage systems on the end-system hosts at these institutions can become a bottleneck for terabit data movement. The parallel storage system (PFS) is shared by simulation systems, experimental systems, analysis and visualization clusters, in addition to wide-area data movers. These competing uses often induce temporary, but significant, I/O load imbalances on the storage system, which impact the performance of all the users. The problem is a serious concern because some resources are more expensive (e.g. super computers) or have time-critical deadlines (e.g. experimental data from a light source), but parallel file systems handle all requests fairly even if some storage servers are under heavy load. This paper investigates the problem of competing workloads accessing the parallel file system and how the performance of wide-area data movement can be improved in these environments. First, we study the I/O load imbalance problems using actual I/O performance data collected from the Spider storage system at the Oak Ridge Leadership Computing Facility. Second, we present I/O optimization solutions with layout-awareness on end-system hosts for bulk data movement. With our evaluation, we show that our I/O optimization techniques can avoid the I/O congested disk groups, improving storage I/O times on parallel storage systems for terabit data movement.

I/O and Storage↗

Automatic Identification of Application I/O Signatures from Noisy Server-Side Traces

Competing workloads on a shared storage system cause I/O resource contention and application performance vagaries. This problem is already evident in today s HPC storage systems and is likely to become acute at exascale. We need more interaction between application I/O requirements and system software tools to help alleviate the I/O bottleneck, moving towards I/O-aware job scheduling. However, this requires rich techniques to capture application I/O characteristics, which remain evasive in production systems. Traditionally, I/O characteristics have been obtained using client-side tracing tools, with drawbacks such as non-trivial instrumentation/development costs, large trace traffic, and inconsistent adoption. We present a novel approach, I/O Signature Identifier (IOSI), to characterize the I/O behavior of data-intensive applications. IOSI extracts signatures from noisy, zero-overhead server-side I/O throughput logs that are already collected on today s supercomputers, without interfering with the compiling/execution of applications. We evaluated IOSI using the Spider storage system at Oak Ridge National Laboratory, the S3D turbulence application (running on 18,000 Titan nodes), and benchmark-based pseudo-applications. Through our ex- periments we confirmed that IOSI effectively extracts an application s I/O signature despite significant server-side noise. Compared to client-side tracing tools, IOSI is transparent, interface-agnostic, and incurs no overhead. Compared to alternative data alignment techniques (e.g., dynamic time warping), it offers higher signature accuracy and shorter processing time.

Liu, Yang↗

DEVELOPMENT OF OPERATIONAL CONCEPTS FOR ADVANCED SMRs: THE ROLE OF COGNITIVE SYSTEMS ENGINEERING

Advanced small modular reactors (AdvSMRs) will use advanced digital instrumentation and control systems, and make greater use of automation. These advances not only pose technical and operational challenges, but will inevitably have an effect on the operating and maintenance (O&M) cost of new plants. However, there is much uncertainty about the impact of AdvSMR designs on operational and human factors considerations, such as workload, situation awareness, human reliability, staffing levels, and the appropriate allocation of functions between the crew and various automated plant systems. Existing human factors and systems engineering design standards and methodologies are not current in terms of human interaction requirements for dynamic automated systems and are no longer suitable for the analysis of evolving operational concepts. New models and guidance for operational concepts for complex socio-technical systems need to adopt a state-of-the-art approach such as Cognitive Systems Engineering (CSE) that gives due consideration to the role of personnel. This approach we report on helps to identify and evaluate human challenges related to non-traditional concepts of operations. A framework - defining operational strategies was developed based on the operational analysis of Argonne National Laboratory’s Experimental Breeder Reactor-II (EBR-II), a small (20MWe) sodium-cooled reactor that was successfully operated for thirty years. Insights from the application of the systematic application of the methodology and its utility are reviewed and arguments for the formal adoption of CSE as a value-added part of the Systems Engineering process are presented.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

SCALING AN URBAN EMERGENCY EVACUATION FRAMEWORK: CHALLENGES AND PRACTICES

Critical infrastructure disruption, caused by severe weather events, natural disasters, terrorist attacks, etc., has significant impacts on urban transportation systems. We built a computational framework to simulate urban transportation systems under critical infrastructure disruption in order to aid real-time emergency evacuation. This framework will use large scale datasets to provide a scalable tool for emergency planning and management. Our framework, World-Wide Emergency Evacuation (WWEE), integrates population distribution and urban infrastructure networks to model travel demand in emergency situations at global level. Also, a computational model of agent-based traffic simulation is used to provide an optimal evacuation plan for traffic operation purpose [1]. In addition, our framework provides a web-based high resolution visualization tool for emergency evacuation modelers and practitioners. We have successfully tested our framework with scenarios in both United States (Alexandria, VA) and Europe (Berlin, Germany) [2]. However, there are still some major drawbacks for scaling this framework to handle big data workloads in real time. On our back-end, lack of proper infrastructure limits us in ability to process large amounts of data, run the simulation efficiently and quickly, and provide fast retrieval and serving of data. On the front-end, the visualization performance of microscopic evacuation results is still not efficient enough due to high volume data communication between server and client. We are addressing these drawbacks by using cloud computing and next-generation web technologies, namely Node.js, NoSQL, WebGL, Open Layers 3 and HTML5 technologies. We will describe briefly about each one and how we are using and leveraging these technologies to provide an efficient tool for emergency management organizations. Our early experimentation demonstrates that using above technologies is a promising approach to build a scalable and high performance urban emergency evacuation framework that can improve traffic mobility and safety under critical infrastructure disruption in today s socially connected world.

cloud computing↗

Scalability Analysis of Gleipnir: A Memory Tracing and Profiling Tool, on Titan

Application performance is hindered by a variety of factors but most notably driven by the well know CPU-memory speed gap (also known as the memory wall). Understanding application s memory behavior is key if we are trying to optimize performance. Understanding application performance properties is facilitated with various performance profiling tools. The scope of profiling tools varies in complexity, ease of deployment, profiling performance, and the detail of profiled information. Specifically, using profiling tools for performance analysis is a common task when optimizing and understanding scientific applications on complex and large scale systems such as Cray s XK7. This paper describes the performance characteristics of using Gleipnir, a memory tracing tool, on the Titan Cray XK7 system when instrumenting large applications such as the Community Earth System Model. Gleipnir is a memory tracing tool built as a plug-in tool for the Valgrind instrumentation framework. The goal of Gleipnir is to provide fine-grained trace information. The generated traces are a stream of executed memory transactions mapped to internal structures per process, thread, function, and finally the data structure or variable. Our focus was to expose tool performance characteristics when using Gleipnir with a combination of an external tools such as a cache simulator, Gl CSim, to characterize the tool s overall performance. In this paper we describe our experience with deploying Gleipnir on the Titan Cray XK7 system, report on the tool s ease-of-use, and analyze run-time performance characteristics under various workloads. While all performance aspects are important we mainly focus on I/O characteristics analysis due to the emphasis on the tools output which are trace-files. Moreover, the tool is dependent on the run-time system to provide the necessary infrastructure to expose low level system detail; therefore, we also discuss any theoretical benefits that can be achieved if such modules were present.

Janjusic, Tommy↗

EqualChance: Addressing Intra-set Write Variation to Increase Lifetime of Non-volatile Caches

To address the limitations of SRAM such as high-leakage and low-density, researchers have explored use of non-volatile memory (NVM) devices, such as ReRAM (resistive RAM) and STT-RAM (spin transfer torque RAM) for designing on-chip caches. A crucial limitation of NVMs, however, is that their write endurance is low and the large intra-set write variation introduced by existing cache management policies may further exacerbate this problem, thereby reducing the cache lifetime significantly. We present EqualChance, a technique to increase cache lifetime by reducing intra-set write variation. EqualChance works by periodically changing the physical cache-block location of a write-intensive data item within a set to achieve wear-leveling. Simulations using workloads from SPEC CPU2006 suite and HPC (high-performance computing) field show that EqualChance improves the cache lifetime by 4.29X. Also, its implementation overhead is small, and it incurs very small performance and energy loss.

cache lifetime↗

The Initial Development of a Computerized Operator Support System

A computerized operator support system (COSS) is a collection of resilient software technologies to assist operators in monitoring overall nuclear power plant performance and making timely, informed decisions on appropriate control actions for the projected plant condition. The COSS provides rapid assessments, computations, and recommendations to reduce workload and augment operator judgment and decision-making during fast- moving, complex events. A prototype COSS for a chemical volume control system at a nuclear power plant has been developed in order to demonstrate the concept and provide a test bed for further research. The development process identified four underlying elements necessary for the prototype, which consist of a digital alarm system, computer-based procedures, piping and instrumentation diagram system representations, and a recommender module for mitigation actions. An operational prototype resides at the Idaho National Laboratory (INL) using the U.S. Department of Energy’s (DOE) Light Water Reactor Sustainability (LWRS) Human Systems Simulation Laboratory (HSSL). Several human-machine interface (HMI) considerations are identified and incorporated in the prototype during this initial round of development.

99 GENERAL AND MISCELLANEOUS↗

The Next Step in Deployment of Computer Based Procedures For Field Workers: Insights And Results From Field Evaluations at Nuclear Power Plants

The paper-based procedures currently used for nearly all activities in the commercial nuclear power industry have a long history of ensuring safe operation of the plants. However, there is potential to greatly increase efficiency and safety by improving how the human operator interacts with the procedures. One way to achieve these improvements is through the use of computer-based procedures (CBPs). A CBP system offers a vast variety of improvements, such as context driven job aids, integrated human performance tools (e.g., placekeeping, correct component verification, etc.), and dynamic step presentation. The latter means that the CBP system could only display relevant steps based on operating mode, plant status, and the task at hand. A dynamic presentation of the procedure (also known as context-sensitive procedures) will guide the operator down the path of relevant steps based on the current conditions. This feature will reduce the operator’s workload and inherently reduce the risk of incorrectly marking a step as not applicable and the risk of incorrectly performing a step that should be marked as not applicable. The research team at the Idaho National Laboratory has developed a prototype CBP system for field workers, which has been evaluated from a human factors and usability perspective in four laboratory studies. Based on the results from each study revisions were made to the CBP system. However, a crucial step to get the end users' (e.g., auxiliary operators, maintenance technicians, etc.) acceptance is to put the system in their hands and let them use it as a part of their everyday work activities. In the spring 2014 the first field evaluation of the INL CBP system was conducted at a nuclear power plant. Auxiliary operators conduct a functional test of one out of three backup air compressors each week. During the field evaluation activity, one auxiliary operator conducted the test with the paper-based procedure while a second auxiliary operator followed along with the computer-based procedure. After each conducted functional test the operators were asked a series of questions designed to provide feedback on the feasibility to use a CBP system in the plant and the general user experience of the CBP system. This paper will describe the field evaluation and its results in detail. For example, the result shows that the context driven job aids and the incorporated human performance tools are much liked by the auxiliary operators. The paper will describe and present initial findings from a second field evaluation conducted at second nuclear utility. For this field evaluation a preventive maintenance work order for the HVAC system was used. In addition, there will be a description of the method and objective of two field evaluations planned to be conducted late 2014 or early 2015.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Particle Communication and Domain Neighbor Coupling: Scalable Domain Decomposed Algorithms for Monte Carlo Particle Transport

In order to run Monte Carlo particle transport calculations on new supercomputers with hundreds of thousands or millions of processors, care must be taken to implement scalable algorithms. This means that the algorithms must continue to perform well as the processor count increases. In this paper, we examine the scalability of:(1) globally resolving the particle locations on the correct processor, (2) deciding that particle streaming communication has finished, and (3) efficiently coupling neighbor domains together with different replication levels. We have run domain decomposed Monte Carlo particle transport on up to 2 21 = 2,097,152 MPI processes on the IBM BG/Q Sequoia supercomputer and observed scalable results that agree with our theoretical predictions. These calculations were carefully constructed to have the same amount of work on every processor, i.e. the calculation is already load balanced. We also examine load imbalanced calculations where each domain’s replication level is proportional to its particle workload. In this case we show how to efficiently couple together adjacent domains to maintain within workgroup load balance and minimize memory usage.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗