Durango: Scalable Synthetic Workloard Generation for Extreme-Scale Application Performance Modeling & Simulation
Explore the source record for details and available documents.
SEARCH · Engineering Papers
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.
Explore the source record for details and available documents.
OpenSHMEM is a community effort to unifyt and standardize the SHMEM programming model. MPI (Message Passing Interface) is a well-known community standard for parallel programming using distributed memory. The most recen t release of MPI, version 3.0, was designed in part to support programming models like SHMEM.OSHMPI is an implementation of the OpenSHMEM standard using MPI-3 for the Linux operating system. It is the first implementation of SHMEM over MPI one-sided communication and has the potential to be widely adopted due to the portability and widely availability of Linux and MPI-3. OSHMPI has been tested on a variety of systems and implementations of MPI-3, includingInfiniBand clusters using MVAPICH2 and SGI shared-memory supercomputers using MPICH. Current support is limited to Linux but may be extended to Apple OSX if there is sufficient interest. The code is opensource via https://github.com/jeffhammond/oshmpi
Extensive computing power has been used to tackle issues such as climate changes, fusion energy, and other pressing scientific challenges. These computations produce a tremendous amount of data; however, many of the data analysis programs currently only run a single processor. In this work, we explore the possibility of using the emerging cloud computing platform to parallelize such sequential data analysis tasks. As a proof of concept, we wrap a program for analyzing trends of tropical cyclones in a set of virtual machines (VMs). This approach allows the user to keep their familiar data analysis environment in the VMs, while we provide the coordination and data transfer services to ensure the necessary input and output are directed to the desired locations. This work extensively exercises the networking capability of the cloud computing systems and has revealed a number of weaknesses in the current cloud system software. In our tests, we are able to scale the parallel data analysis job to a modest number of VMs and achieve a speedup that is comparable to running the same analysis task using MPI. However, compared to MPI based parallelization, the cloud-based approach has a number of advantages. The cloud-based approach is more flexible because the VMs can capture arbitrary software dependencies without requiring the user to rewrite their programs. The cloud-based approach is also more resilient to failure; as long as a single VM is running, it can make progress while as soon as one MPI node fails the whole analysis job fails. In short, this initial work demonstrates that a cloud computing system is a viable platform for distributed scientific data analyses traditionally conducted on dedicated supercomputing systems.
OpenSHMEM is a community effort to unifyt and standardize the SHMEM programming model. MPI (Message Passing Interface) is a well-known community standard for parallel programming using distributed memory. The most recen t release of MPI, version 3.0, was designed in part to support programming models like SHMEM.OSHMPI is an implementation of the OpenSHMEM standard using MPI-3 for the Linux operating system. It is the first implementation of SHMEM over MPI one-sided communication and has the potential to be widely adopted due to the portability and widely availability of Linux and MPI-3. OSHMPI has been tested on a variety of systems and implementations of MPI-3, includingInfiniBand clusters using MVAPICH2 and SGI shared-memory supercomputers using MPICH. Current support is limited to Linux but may be extended to Apple OSX if there is sufficient interest. The code is opensource via https://github.com/jeffhammond/oshmpi
Important insights into many problems that are traditionally analyzed via statistical models can be obtained by re-formulating and evaluating within a large-scale optimization framework. The theoretical underpinnings of the statistical model often shift the goal of the solution space traversal from a traditional search for a single optimal solution to a traversal with the purpose of yielding a set of high quality, independent solutions. We examine statistical frameworks with astronomical solution spaces where the independence requirement constitutes a significant additional challenge for standard optimization methodologies. We design a hybrid metaheuristic with specialized intensification and diversification protocols in the base search algorithm. We extend our algorithm to the high-performance-computing realm using the Stampede2 supercomputer. We experimentally demonstrate the effectiveness of our algorithm to utilize multiple processors to collaboratively hill climb, broadcast messages to one another regarding landscape characteristics, diversify across the solution landscape, and request aid in climbing particularly difficult peaks.
Mesoscopic simulations of hydrocarbon flow in source shales are concerning, in part due to the heterogeneous shale pores with sizes ranging from a few nanometers to a few micrometers. Additionally, the sub-continuum fluid–fluid and fluid–solid interactions in nano- to micro-scale shale pores, which are physically and chemically sophisticated, must be captured. To address those challenges, we present a GPU-accelerated package for simulation of flow in nano- to micro-pore networks with a many-body dissipative particle dynamics (mDPD) mesoscale model. Based on a fully distributed parallel paradigm, the code offloads all intensive workloads on GPUs. Other advancements, such as smart particle packing and no-slip boundary condition in complex pore geometries, are also implemented for the construction and the simulation of the realistic shale pores from 3D nanometer-resolution stack images. Our code is validated for accuracy and compared against the CPU counterpart for speedup. In our benchmark tests, the code delivers nearly perfect strong scaling and weak scaling (with up to 512 million particles) on up to 512 K20X GPUs on Oak Ridge National Laboratory’s (ORNL) Titan supercomputer. Moreover, a single-GPU benchmark on ORNL’s SummitDev and IBM’s AC922 suggests that the host-to-device NVLink can boost performance over PCIe by a remarkable 40%. Lastly, we demonstrate, through a flow simulation in realistic shale pores, that the CPU counterpart requires 840 Power9 cores to rival the performance delivered by our package with four V100 GPUs on ORNL’s Summit architecture. This simulation package enables quick-turnaround and high-throughput mesoscopic numerical simulations for exploring complex flow phenomena in nano- to micro-porous rocks with realistic pore geometries.
We introduce and study numerically a scalable parallel finite element solver for the simulation of blood flow in compliant arteries. The incompressible Navier-Stokes equations are used to model the fluid and coupled to an incompressible linear elastic model for the blood vessel walls. Our method features an unstructured dynamic mesh capable of modeling complicated geometries, an arbitrary Lagrangian-Eulerian framework that allows for large displacements of the moving fluid domain, monolithic coupling between the fluid and structure equations, and fully implicit time discretization. Simulations based on blood vessel geometries derived from patient-specific clinical data are performed on large supercomputers using scalable Newton-Krylov algorithms preconditioned with an overlapping restricted additive Schwarz method that preconditions the entire fluid-structure system together. The algorithm is shown to be robust and scalable for a variety of physical parameters, scaling to hundreds of processors and millions of unknowns.
Here we present a highly scalable, parallel implementation of first-principles electron dynamics coupled with molecular dynamics (MD). By using optimized kernels, network topology aware communication, and by fully distributing all terms in the time-dependent Kohn–Sham equation, we demonstrate unprecedented time to solution for disordered aluminum systems of 2000 atoms (22,000 electrons) and 5400 atoms (59,400 electrons), with wall clock time as low as 7.5 s per MD time step. Despite a significant amount of non-local communication required in every iteration, we achieved excellent strong scaling and sustained performance on the Sequoia Blue Gene/Q supercomputer at LLNL. We obtained up to 59% of the theoretical sustained peak performance on 16,384 nodes and performance of 8.75 Petaflop/s (43% of theoretical peak) on the full 98,304 node machine (1,572,864 cores). Lastly, scalable explicit electron dynamics allows for the study of phenomena beyond the reach of standard first-principles MD, in particular, materials subject to strong or rapid perturbations, such as pulsed electromagnetic radiation, particle irradiation, or strong electric currents.
The simulation of multiscale physics is an important challenge for scientific computing. For this class of problem, large three-dimensional simulations are performed to advance scientific inquiry. On massively parallel computing systems, the volume of data generated by such approaches can become a productivity bottleneck if the raw data generated from the simulation is analyzed in a post-processing step. To address this, we present a physics-based framework for in situ data reduction that is theoretically grounded in multiscale averaging theory. We show how task parallelism can be exploited to concurrently perform a variety of analysis tasks with data-dependent costs, including the generation of iso-surfaces, morphological analyses, and connected components analysis. All analyses are performed in parallel using distributed memory and use the same domain decomposition as the simulation. A task management framework is constructed to leverage available parallelism within a node for analysis. The capabilities of the framework are to launch asynchronous analysis threads, manage dependencies between different tasks, promote data locality and minimize the impact of data transfers. The framework is applied to analyze GPU-based simulations of two-fluid-phase flow in porous media, generating a set of averaged measures that represents the overall system behavior. We demonstrate how the approach can be applied to perform physically-consistent analysis over fluid sub-regions determined from connected components analysis. Simulations performed on Oak Ridge National Lab's Titan supercomputer are profiled to demonstrate the performance of the associated multi-threaded in situ analysis approach for typical production simulation of two-fluid-phase flow.
The US Department of Energy (DOE) conducts climate simulations on some of the world’s largest supercomputers. These exascale machines use heterogeneous architectures with both CPUs and GPUs, and scientific codes must adapt to make full use of this computing power. Los Alamos National Lab is developing Omega: The Ocean Model for E3SM Global Applications, which is specifically designed for modern exascale computers. It uses external libraries that have been optimized for a variety of architectures to run on different supercomputers. Omega is an unstructured-mesh ocean model based on TRiSK numerical methods. It will be the new ocean component of the DOE’s Energy Exascale Earth System Model (E3SM). The algorithms in Omega follow those of the current ocean component, MPAS-Ocean, but it will be written in C++ rather than Fortran to take advantage of the Kokkos performance portability library. Omega spatial operators are written as Kokkos kernels to run efficiently on both CPUs and GPUs. Work on Omega began in 2023 with a new C++ framework for unstructured mesh partitioning, halo exchanges, parallel IO, and Kokkos interfaces. The current version, Omega-0, is being developed to solve the shallow water equations and at present includes all of the tendency terms but not time stepping. Here we share the results of Omega-0 verification and performance testing. Verification includes unit tests implemented with CTest as well as convergence tests in Polaris, an in-house python package with a large suite of test problems. Performance tests compare simulations conducted on CPUs versus GPUs and across different architectures: tests are run on Frontier, which has AMD “Optimized 3rd Gen EPYC” CPUs and AMD MI250X GPUs, as well as Perlmutter, which is composed of AMD EPYC 7763 CPUs and NVIDIA A100 GPUs.
Scientific analyses commonly compose multiple single-process programs into a dataflow. An end-to-end dataflow of single-process programs is known as a many-task application. Typically, HPC tools are used to parallelize these analyses. In this work, we investigate an alternate approach that uses Apache Spark-a modern platform for data intensive computing-to parallelize many-task applications. We implement Kira, a flexible and distributed astronomy image processing toolkit, and its Source Extractor (Kira SE) application. Using Kira SE as a case study, we examine the programming flexibility, dataflow richness, scheduling capacity and performance of Apache Spark running on the Amazon EC2 cloud. By exploiting data locality, Kira SE achieves a 4.1× speedup over an equivalent C program when analyzing a 1TB dataset using 512 cores on the Amazon EC2 cloud. Furthermore, Kira SE on the Amazon EC2 cloud achieves a 1.8× speedup over the C program on the NERSC Edison supercomputer. A 128-core Amazon EC2 cloud deployment of Kira SE using Spark Streaming can achieve a second-scale latency with a sustained throughput of 800 MB/s. Our experience with Kira demonstrates that data intensive computing platforms like Apache Spark are a performant alternative for many-task scientific applications.
Leading high performance computing systems achieve their status through use of highly parallel devices such as NVIDIA graphics processing units or Intel Xeon Phi many-core CPUs. The concept of performance portability across such architectures, as well as traditional CPUs, is vital for the application programmer. In this paper we describe targetDP, a lightweight abstraction layer which allows grid-based applications to target data parallel hardware in a platform agnostic manner. We demonstrate the effectiveness of our pragmatic approach by presenting performance results for a complex fluid application (with which the model was co-designed), plus separate lattice quantum chromodynamics particle physics code. For each application, a single source code base is seen to achieve portable performance, as assessed within the context of the Roofline model. TargetDP can be combined with Message Passing Interface (MPI) to allow use on systems containing multiple nodes: we demonstrate this through provision of scaling results on traditional and graphics processing unit-accelerated large scale supercomputers.
This work develops new techniques within Horovod, a generic communication library supporting data parallel training across deep learning frameworks. In particular, we improve the Horovod control plane by implementing a new coordination scheme that takes advantage of the characteristics of the typical data parallel training paradigm, namely the repeated execution of collectives on the gradients of a fixed set of tensors. Using a caching strategy, we execute Horovod’s existing coordinator-worker logic only once during a typical training run, replacing it with a more efficient decentralized orchestration strategy using the cached data and a global intersection of a bitvector for the remaining training duration. Next, we introduce a feature for end users to explicitly group collective operations, enabling finer grained control over the communication buffer sizes. To evaluate our proposed strategies, we conduct experiments on a world-class supercomputer — Summit. We compare our proposals to Horovod’s original design and observe 2x performance improvement at a scale of 6000 GPUs; we also compare them against tf.distribute and torch.DDP and achieve 12% better and comparable performance, respectively, using up to 1536 GPUs; we compare our solution against BytePS in typical HPC settings and achieve about 20% better performance on a scale of 768 GPUs. Finally, we test our strategies on a scientific application (STEMDL) using up to 27,600 GPUs (the entire Summit) and show that we achieve a near-linear scaling of 0.93 with a sustained performance of 1.54 exaflops (with standard error +- 0.02) in FP16 precision.
Patch-based structured adaptive mesh refinement (SAMR) is widely used for high-resolution simu- lations. Combined with modern supercomputers, it could provide simulations of unprecedented size and resolution. A persistent challenge for this com- bination has been managing dynamically adaptive meshes on more and more MPI tasks. The dis- tributed mesh management scheme in SAMRAI has made some progress SAMR scalability, but early al- gorithms still had trouble scaling past the regime of 105 MPI tasks. This work provides two critical SAMR regridding algorithms, which are integrated into that scheme to ensure efficiency of the whole. The clustering algorithm is an extension of the tile- clustering approach, making it more flexible and efficient in both clustering and parallelism. The partitioner is a new algorithm designed to prevent the network congestion experienced by its prede- cessor. We evaluated performance using weak- and strong-scaling benchmarks designed to be difficult for dynamic adaptivity. Results show good scaling on up to 1.5M cores and 2M MPI tasks. Detailed timing diagnostics suggest scaling would continue well past that.
Performance-portable programming frameworks provide abstractions for parallel execution to allow easily porting an application to multiple backend programming models, such as CUDA, HIP, and OpenMP. However, programs may still have portability bugs that manifest only on specific backends. Traditional testing is ineffective in discovering these bugs, as it would require concrete execution on all supported hardware configurations for a potentially infinite set of inputs. To mitigate this issue, we focused on a specific programming framework, Kokkos, and identified several categories of common portability bugs. We then developed Klokkos, a static analysis approach based on symbolic execution that can run on commodity hardware, before execution on supercomputers. As a proof-of-concept, we ran Klokkos on examples encoding the identified bugs. Our results show that Klokkos is effective, efficient, and precise: it detected all the considered bugs, quickly, and without any false positives. Although preliminary, our results motivate further research and development in this direction.
The design of HPC file and storage systems has largely been driven by the requirements on capability, reliability, and capacity. However, the convergence of large-scale simulations with big data analytics have put the data, its usability, and management back on the front and center position. One of the most common and time consuming data management tasks is the transfer of very large datasets within and across file systems.In this paper, we are introducing the FCP tool, a file system agnostic copy tool designed at the OLCF for scalable and high-performance data transfers between two file system endpoints. It provides an array of interesting features such as adaptive chunking, checksumming on the fly, checkpoint and resume capabilities to handle failures, and preserving stripe information for Lustre file system among others. It is currently available on the Titan supercomputer at the OLCF. Initial tests have shown that FCP has much better and scalable performance than traditional data copy tools and it was capable of transferring petabyte-scale datasets between two Lustre file systems.
Orchestrating the execution of ensembles of processes lies at the core of scientific workflow engines on large scale parallel platforms. This is usually handled using platform-specific command line tools, with limited process management control and potential strain on system resources. The PMIx standard provides a uniform interface to system resources. The low level C implementation of PMIx has hampered its use in workflow engines, leading to the development of Python binding that has yet to gain traction. In this paper, we present our work to harden the PMIx Python client, demonstrating its usability using a prototype Python driver to orchestrate the execution of an ensemble of processes. We present experimental results using the prototype on the Summit supercomputer at Oak Ridge National Laboratory. This work lays the foundation for wider adoption of PMIx for workflow engines, and encourages wider support of more PMIx functionality in vendor provided system software stacks.
The primary objective of this research is to support the design of liquid rocket systems for the Advanced Space Transportation System. Since the space launch systems in the near future are likely to rely on liquid rocket engines, increasing the efficiency and reliability of the engine components is an important task. One of the major problems in the liquid rocket engine is to understand fluid dynamics of fuel and oxidizer flows from the fuel tank to plume. Understanding the flow through the entire turbo-pump geometry through numerical simulation will be of significant value toward design. One of the milestones of this effort is to develop, apply and demonstrate the capability and accuracy of 3D CFD methods as efficient design analysis tools on high performance computer platforms. The development of the Message Passage Interface (MPI) and Multi Level Parallel (MLP) versions of the INS3D code is currently underway. The serial version of INS3D code is a multidimensional incompressible Navier-Stokes solver based on overset grid technology, INS3D-MPI is based on the explicit massage-passing interface across processors and is primarily suited for distributed memory systems. INS3D-MLP is based on multi-level parallel method and is suitable for distributed-shared memory systems. For the entire turbo-pump simulations, moving boundary capability and efficient time-accurate integration methods are built in the flow solver, To handle the geometric complexity and moving boundary problems, an overset grid scheme is incorporated with the solver so that new connectivity data will be obtained at each time step. The Chimera overlapped grid scheme allows subdomains move relative to each other, and provides a great flexibility when the boundary movement creates large displacements. Two numerical procedures, one based on artificial compressibility method and the other pressure projection method, are outlined for obtaining time-accurate solutions of the incompressible Navier-Stokes equations. The performance of the two methods is compared by obtaining unsteady solutions for the evolution of twin vortices behind a flat plate. Calculated results are compared with experimental and other numerical results. For an unsteady flow, which requires small physical time step, the pressure projection method was found to be computationally efficient since it does not require any subiteration procedure. It was observed that the artificial compressibility method requires a fast convergence scheme at each physical time step in order to satisfy the incompressibility condition. This was obtained by using a GMRES-ILU(0) solver in present computations. When a line-relaxation scheme was used, the time accuracy was degraded and time-accurate computations became very expensive.