Search NASA⌕ Search

SEARCH · Search NASA

Results for “HPC system”

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 253 records · Page 14

Power/energy use cases for high performance computing

Power and Energy have been identified as a first order challenge for future extreme scale high performance computing (HPC) systems. In practice the breakthroughs will need to be provided by the hardware vendors. But to make the best use of the solutions in an HPC environment, it will likely require periodic tuning by facility operators and software components. This document describes the actions and interactions needed to maximize power resources. It strives to cover the entire operational space in which an HPC system occupies. The descriptions are presented as formal use cases, as documented in the Unified Modeling Language Specification [1]. The document is intended to provide a common understanding to the HPC community of the necessary management and control capabilities. Assuming a common understanding can be achieved, the next step will be to develop a set of Application Programing Interfaces (APIs) to which hardware vendors and software developers could utilize to steer power consumption.

97 MATHEMATICS AND COMPUTING↗

Supporting multiple hardware architectures at CMS: the integration and validation of POWER9

Computing resources in the Worldwide LHC Computing Grid (WLCG) have been based entirely on the x86 architecture for more than two decades. In the near future, however, heterogeneous non-x86 resources, such as ARM, POWER and Risc-V, will become a substantial fraction of the resources that will be provided to the LHC experiments, due to their presence in existing and planned world-class HPC installations. The CMS experiment, one of the four large detectors at the LHC, has started to prepare for this situation, with the CMS software stack (CMSSW) already compiled for multiple architectures. In order to allow for a production use, the tools for workload management and job distribution need to be extended to be able to exploit heterogeneous architectures. Profiting from the opportunity to exploit the first sizable IBM Power9 allocation available on Marconi100 HPC system at CINECA, CMS developed all the needed modifications to the CMS workload management system. After a successful proof of concept, a full physics validation has been performed in order to bring the system in production. The experiences are of very high value, when it comes to commissioning of the similar (even larger) Summit HPC system at Oak Ridge, where CMS is also expecting a resource allocation. Moreover the compute power of those systems is being provided also via GPUs and this represents an extremely valuable opportunity to exploit the offloading capability already implemented in CMSSW. The status of the current integration including the exploitation of the GPUs, the results of the validation as well as the future plans will be shown and discussed.

Boccali, Tommaso [INFN, Pisa]↗

Continuous whole-system monitoring toward rapid understanding of production HPC applications and systems

A detailed understanding of HPC applications’ resource needs and their complex interactions with each other and HPC platform resources are critical to achieving scalability and performance. Such understanding has been difficult to achieve because typical application profiling tools do not capture the behaviors of codes under the potentially wide spectrum of actual production conditions and because typical monitoring tools do not capture system resource usage information with high enough fidelity to gain sufficient insight into application performance and demands. In this paper we present both system and application profiling results based on data obtained through synchronized system wide monitoring on a production HPC cluster at Sandia National Laboratories (SNL). We demonstrate analytic and visualization techniques that we are using to characterize application and system resource usage under production conditions for better understanding of application resource needs. Furthermore, our goals are to improve application performance (through understanding application-to-resource mapping and system throughput) and to ensure that future system capabilities match their intended workloads.

97 MATHEMATICS AND COMPUTING↗

Adaptive Learning for Concept Drift in Application Performance Modeling

Supervised learning is a promising approach for modeling the performance of applications running on large HPC systems. A key assumption in supervised learning is that the training and testing data are obtained under the same conditions. However, in production HPC systems these conditions might not hold because the conditions of the platform can change over time as a result of hardware degradation, hardware replacement, software upgrade, and configuration updates. These changes could alter the data distribution in a way that affects the accuracy of the predictive performance models and render them less useful; this phenomenon is referred to as concept drift. Ignoring concept drift can lead to suboptimal resource usage and decreased efficiency when those performance models are deployed for tuning and job scheduling in production systems. To address this issue, we propose a concept-drift-aware predictive modeling approach that comprises two components: (1) an online Bayesian changepoint detection method that can automatically identify the location of events that lead to concept drift in near-real time and (2) a moment-matching transformation inspired by transfer learning that converts the training data collected before the drift to be useful for retraining. We use application input/output performance data collected on Cori, a production supercomputing system at the National Energy Research Scientific Computing Center, to demonstrate the effectiveness of our approach. The results show that concept-drift-aware models obtain significant improvement in accuracy; the median absolute error of the best-performing Gaussian process regression improved by 58.8% when the proposed approaches were used.

adaptive learning↗

Designing and prototyping extensions to the Message Passing Interface in MPICH

As HPC system architectures and the applications running on them continue to evolve, the MPI standard itself must evolve. The trend in current and future HPC systems toward powerful nodes with multiple CPU cores and multiple GPU accelerators makes efficient support for hybrid programming critical for applications to achieve high performance. However, the support for hybrid programming in the MPI standard has not kept up with recent trends. The MPICH implementation of MPI provides a platform for implementing and experimenting with new proposals and extensions to fill this gap and to gain valuable experience and feedback before the MPI Forum can consider them for standardization. Here, in this work, we detail six extensions implemented in MPICH to increase MPI interoperability with other runtimes, with a specific focus on heterogeneous architectures. First, the extension to MPI generalized requests lets applications integrate asynchronous tasks into MPI’s progress engine. Second, the iovec extension to datatypes lets applications use MPI datatypes as a general-purpose data layout API beyond just MPI communications. Third, a new MPI object, MPIX_Stream, can be used by applications to identify execution contexts beyond MPI processes, including threads and GPU streams. MPIX stream communicators can be created to make existing MPI functions thread-aware and GPU-aware, thus providing applications with explicit ways to achieve higher performance. Fourth, MPIX Streams are extended to support the enqueue semantics for offloading MPI communications onto a GPU stream context. Fifth, thread communicators allow MPI communicators to be constructed with individual threads, thus providing a new level of interoperability between MPI and on-node runtimes such as OpenMP. Lastly, we present an extension to invoke MPI progress, which lets users spawn progress threads with fine-grained control to adapt the communication performance to their application designs. We describe the design and implementation of these extensions, provide usage examples, and highlight their expected benefits with performance results.

97 MATHEMATICS AND COMPUTING↗

ZFP Hardware Implementation

As core counts increase in new HPC systems with comparatively little increase in memory bandwidth, the trend is an effective decrease in memory bandwidth per core. Other bandwidth limitations in HPC systems exist between CPU and GPU memory, between system nodes, and between node memory and storage. Compression of floating-point data has the potential to reduce data movement and pressure across these communication channels. Furthermore, it has the potential to reduce the footprint of floating-point arrays stored in memory. ZFP, implemented in software, is gaining traction as an effective method in floating-point compression; however, performance gains are limited to the spare compute cycles available before reaching the bandwidth limitations of the communication channel. A hardware implementation of ZFP has the potential to raise the bar on performance. From the inception of ZFP, it was designed to accommodate a hardware implementation.

42 ENGINEERING↗

Optimization and Portability of a Fusion OpenACC-based FORTRAN HPC Code from NVIDIA to AMD GPUs

NVIDIA has been the main provider of GPU hardware in HPC systems for over a decade. Most applications that benefit from GPUs have thus been developed and optimized for the NVIDIA software stack. Recent exascale HPC systems are, however, introducing GPUs from other vendors, e.g. with the AMD GPU-based OLCF Frontier system just becoming available. AMD GPUs cannot be directly accessed using the NVIDIA software stack, and require a porting effort by the application developers. This paper provides an overview of our experience porting and optimizing the CGYRO code, a widely-used fusion simulation tool based on FORTRAN with OpenACC-based GPU acceleration. While the porting from the NVIDIA compilers was relatively straightforward using the CRAY compilers on the AMD systems, the performance optimization required more fine-tuning. In the optimization effort, we uncovered code sections that had performed well on NVIDIA GPUs, but were unexpectedly slow on AMD GPUs. After AMD-targeted code optimizations, performance on AMD GPUs has increased to meet our expectations. Modest speed improvements were also seen on NVIDIA GPUs, which was an unexpected benefit of this exercise.

Sfiligoi, Igor↗

Scaling the memory wall using mixed-precision - HPG-MxP on an exascale-class machine

Mixed-precision algorithms have been proposed as a way for scientific computing to benefit from some of the gains seen for AI on recent high performance computing (HPC) platforms. A few applications dominated by dense matrix operations have seen substantial speedups by utilizing low precision formats such as FP16. However, a majority of scientific simulation applications are memory bandwidth limited. Beyond preliminary studies, the practical gain from using mixed-precision algorithms on a given high-performance computing (HPC) system is largely unclear. The High Performance GMRES Mixed Precision (HPG-MxP) benchmark has been proposed to measure the useful performance of a HPC system on sparse matrix-based mixed-precision applications. In this work, we present an implementation of the HPG-MxP benchmark for an exascale system and describe our algorithm enhancements. We show for the first time a speedup of 1.6x using a combination of double- and single-precision keeping the same residual level on modern GPU-based supercomputers.

Kashi, Aditya [ORNL] (ORCID:0000000325893792)↗

On the Impact of Execution Models: A Case Study in Computational Chemistry

Efficient utilization of high-performance computing (HPC) platforms is an important and complex problem. Execution models, abstract descriptions of the dynamic runtime behavior of the execution stack, have significant impact on the utilization of HPC systems. Using a computational chemistry kernel as a case study and a wide variety of execution models combined with load balancing techniques, we explore the impact of execution models on the utilization of an HPC system. We demonstrate a 50 percent improvement in performance by using work stealing relative to a more traditional static scheduling approach. We also use a novel semi-matching technique for load balancing that has comparable performance to a traditional hypergraph-based partitioning implementation, which is computationally expensive. Using this study, we found that execution model design choices and assumptions can limit critical optimizations such as global, dynamic load balancing and finding the correct balance between available work units and different system and runtime overheads. With the emergence of multi- and many-core architectures and the consequent growth in the complexity of HPC platforms, we believe that these lessons will be beneficial to researchers tuning diverse applications on modern HPC platforms, especially on emerging dynamic platforms with energy-induced performance variability.

Execution models, computational chemistry, MPI and↗

Evaluating the Shared Root File System Approach for Diskless High-Performance Computing Systems

Diskless high-performance computing (HPC) systems utilizing networked storage have become popular in the last several years. Removing disk drives significantly increases compute node reliability as they are known to be a major source of failures. Furthermore, networked storage solutions utilizing parallel I/O and replication are able to provide increased scalability and availability. Reducing a compute node to processor(s), memory and network interface(s) greatly reduces its physical size, which in turn allows for large-scale dense HPC solutions. However, one major obstacle is the requirement by certain operating systems (OSs), such as Linux, for a root file system. While one solution is to remove this requirement from the OS, another is to share the root file system over the networked storage. This paper evaluates three networked file system solutions, NFSv4, Lustre and PVFS2, with respect to their performance, scalability, and availability features for servicing a common root file system in a diskless HPC configuration. Our findings indicate that Lustre is a viable solution as it meets both, scaling and performance requirements. However, certain availability issues regarding single points of failure and control need to be considered.

97 MATHEMATICS AND COMPUTING↗

OLCF Test Harness

Acceptance and regression testing of a High Performance Computing (HPC) system requires an automated and reproducible framework and tool for running and logging results. Manually running tests across a system is labor intensive and prone to reproducibility errors. The OLCF Test Harness (OTH) provides a framework in which to document required tests for a HPC system. The OTH then provides tools to execute and log results of these tests in an automated fashion.

Dietz, Dan [Oak Ridge National Laboratory (ORNL), ↗

User-based I/O Profiling for Leadership Scale HPC Workloads

I/O constitutes a significant portion of most of the application run-time. Spawning many such applications concurrently on an HPC system leads to severe I/O contention. Thus, understanding and subsequently reducing I/O contention induced by such multi-tenancy is critical for the efficient and reliable performance of the HPC system. In this study, we demonstrate that an application’s performance is influenced by the command line arguments passed to the job submission. We model an application’s I/O behavior based on two factors: past I/O behavior within a time window and user-configured I/O settings via command-line arguments. We conclude that I/O patterns for well-known HPC applications like E3SM and LAMMP are predictable, with an average uncertainty below 0.25 (A probability of 80%) and near zero (A probability of 100%) within a day. However, I/O pattern variance increases as the study time window lengthens. Additionally, we show that for 38 users and at least 50 applications constituting approximately 93000 job submissions, there is a high correlation between a submitted command line and the past command lines made within 1 to 10 days submitted by the user. We claim the length of this time window is unique per user.

Yazdani, Ahmad Hossein [Virginia Tech, Blacksburg]↗

Software Quality Assurance for High Performance Computing Containers

Software containers are a key channel for delivering portable and reproducible scientific software in high performance computing (HPC) environments. HPC environments are different from other types of computing environments primarily due to usage of the message passing interface (MPI) and drivers for specialized hard- ware to enable distributed computing capabilities. This distinction directly impacts how software containers are built for HPC applications and can complicate software quality assurance efforts including portability and performance. This work introduces a strategy for building containers for HPC applications that adopts layering as a mechanism for software quality assurance. The strategy is demonstrated across three different HPC systems, two of them petaflops scale with entirely different interconnect technologies and/or processor chipsets but running the same container. Performance consequences of the containerization strategy are found to be less than 5-14% while still achieving portable and reproducible containers for HPC systems.

97 MATHEMATICS AND COMPUTING↗

Investigating the Interplay between Energy Efficiency and Resilience in High Performance Computing

Energy efficiency and resilience are two crucial challenges for HPC systems to reach exascale. While energy efficiency and resilience issues have been extensively studied individually, little has been done to understand the interplay between energy efficiency and resilience for HPC systems. Decreasing the supply voltage associated with a given operating frequency for processors and other CMOS-based components can significantly reduce power consumption. However, this often raises system failure rates and consequently increases application execution time. In this work, we present an energy saving undervolting approach that leverages the mainstream resilience techniques to tolerate the increased failures caused by undervolting.

failures↗

Job Scheduler-Driven Power Gateway for High Performance Computing

Power gateways in the form of a microgrid can incorporate multiple distributed energy resources (DER) in either grid forming or grid following mode and support high performance computing (HPC) power profiles including the large load-follow requirements observed in multi-user HPC systems. The microgrid’s flexibility to operate in either grid forming or grid following mode and to actively switch between these modes enables baseline power from multiple non-baseline DER while maintaining high power quality metrics for the HPC system. But this enormous flexibility in demand response and time of use shifting is generally programmed independently of any integration with an HPC job scheduler which can better inform the load shaping by the microgrid. While there are many existing approaches where the HPC job scheduler takes in information from the grid to make queue scheduling decisions, this work takes the opposite view and explores a scheduler where the jobs in the queue can directly impact the settings of the grid. Several HPC scheduler strategies are tested where the jobs in the queue directly impact the settings of a microgrid designed for HPC operation which is driving a datacenter with three classes of HPC architectures. The scheduler operation is shown using a microgrid with 64 kW of solar capacity and 320 kWh of battery over a period of 21 days operating with significant low-follow swings, a throttled grid, cloudy conditions, switching between grid following and grid forming modes, and a wide range of battery states-of-charge all while maintaining high quality power metrics. The scheduler provides a mechanism for the job queue to directly impact a power gateway like a microgrid and to improve HPC power outcomes such as maximizing renewable energy usage

microgrid↗

Detailed Modeling and Evaluation of a Scalable Multilevel Checkpointing System

High-performance computing (HPC) systems are growing more powerful by utilizing more components. As the system mean time before failure correspondingly drops, applications must checkpoint frequently to make progress. But, at scale, the cost of checkpointing becomes prohibitive. A solution to this problem is multilevel checkpointing, which employs multiple types of checkpoints in a single run. Moreover, lightweight checkpoints can handle the most common failure modes, while more expensive checkpoints can handle severe failures. We designed a multilevel checkpointing library, the Scalable Checkpoint/Restart (SCR) library, that writes lightweight checkpoints to node-local storage in addition to the parallel file system. We present probabilistic Markov models of SCR's performance. We show that on future large-scale systems, SCR can lead to a gain in machine efficiency of up to 35 percent, and reduce the load on the parallel file system by a factor of two. In addition, we predict that checkpoint scavenging, or only writing checkpoints to the parallel file system on application termination, can reduce the load on the parallel file system by 20 × on today's systems and still maintain high application efficiency.

97 MATHEMATICS AND COMPUTING↗

Detailed Modeling, Design, and Evaluation of a Scalable Multi-level Checkpointing System

High-performance computing (HPC) systems are growing more powerful by utilizing more hardware components. As the system mean-time-before-failure correspondingly drops, applications must checkpoint more frequently to make progress. However, as the system memory sizes grow faster than the bandwidth to the parallel file system, the cost of checkpointing begins to dominate application run times. A potential solution to this problem is to use multi-level checkpointing, which employs multiple types of checkpoints with different costs and different levels of resiliency in a single run. The goal is to design light-weight checkpoints to handle the most common failure modes and rely on more expensive checkpoints for less common, but more severe failures. While this approach is theoretically promising, it has not been fully evaluated in a large-scale, production system context. To this end we have designed a system, called the Scalable Checkpoint/Restart (SCR) library, that writes checkpoints to storage on the compute nodes utilizing RAM, Flash, or disk, in addition to the parallel file system. We present the performance and reliability properties of SCR as well as a probabilistic Markov model that predicts its performance on current and future systems. We show that multi-level checkpointing improves efficiency on existing large-scale systems and that this benefit increases as the system size grows. In particular, we developed low-cost checkpoint schemes that are 100x-1000x faster than the parallel file system and effective against 85% of our system failures. This leads to a gain in machine efficiency of up to 35%, and it reduces the the load on the parallel file system by a factor of two on current and future systems.

99 GENERAL AND MISCELLANEOUS↗

Fault Oblivious eXascale Whitepaper

In this paper we present a software system which supports dynamic, irregular, adaptive applications. Data objects are created and structured in a hierarchical manner, with replication as needed to provide a high degree of redundancy. The data objects can contain data, code, tasks (work descriptors with references to data, code, and other tasks) and higher level structures such as work queues. The higher level structures benefit from the properties of the data objects: redundant storage to support resiliency in the face of hardware failure; hierarchical structure to optimize use of the HPC system; and a presence of object names, available in the per-user file system name space, which allows any application, not just specially written HPC applications, to make use of the data even while it is on the HPC system. Our use of hierarchy will make the runtime scalable to very large systems. Our use of redundancy will allow programs to be written in a fault-oblivious manner, eliminating the need for system-level checkpointing. Putting data object names into the file system name space allows for interactive use of the system by users. With this approach, we will be able to finally leave the batch era behind, a half-century after the invention of time sharing. We will be able to stop bounding program through- put by the checkpoint interval. Application data will be accessible at any time, not hidden behind opaque 128-bit pointers or MPI ranks, but given a name that is visible everywhere. Programmers can stop laying out data, and thinking about where the data is, and the code is, and the nodes are, and stick with the problem of what the application is supposed to be doing. This work, if it succeeds, will enable scientific computing to scale to the next generation of machines.

99 GENERAL AND MISCELLANEOUS↗