Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Computational Load Balancing”

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 19 records

Massive Contingency Analysis with High Performance Computing

Contingency analysis is a key function in the Energy Management System (EMS) to assess the impact of various combinations of power system component failures based on state estimates. Contingency analysis is also extensively used in power market operation for feasibility test of market solutions. Faster analysis of more cases is required to safely and reliably operate today’s power grids with less marginal and more intermittent renewable energy sources. Enabled by the latest development in the computer industry, high performance computing holds the promise of meet the need in the power industry. This paper investigates the potential of high performance computing for massive contingency analysis. The framework of "N-x" contingency analysis is established and computational load balancing schemes are studied and implemented with high performance computers. Case studies of massive 300,000-contingency-case analysis using the Western Electricity Coordinating Council power grid model are presented to illustrate the application of high performance computing and demonstrate the performance of the framework and computational load balancing schemes.

Computational Load Balancing↗

EJFAT: Towards Intelligent Compute Destination Load Balancing

To handle increased data flow, Jefferson Lab (JLab) is partnering with ESnet for development of an AI/ML directed compute work Load Balancer (LB) of UDP streamed data. The LB is FPGA based featuring dynamically configurable, low latency and high throughput destination address switching. The LB provides integration of edge and core computing to support JLab experimental programs, the Electron-Ion Collider, as well as data centers of the future. In the ESnet/JLab FPGA Accelerated Transport (EJFAT) initiative, the function of the LB Data Plane (DP) is to redirect data streams to selectable (but unknown to sender) destination hosts based on current worload and within that host to destination ports as a function of sub- stream id. This effects hierarchical scaling, first across compute machines for processing over a series of events and second, across ports so different data source sub-streams may be assigned to different processors for further parallelization. The LB Control Plane (CP) programs the DP using compute farm telemetry to direct and balance workloads across a compute cluster as the operating conditions require. While Proportional/Integrative/Derivative (PID) controllers are often seen in similar applications, here we investigate the feasibility of a Reinforcement Learning (RL) based schedule manager running in the CP to provide dynamic updates to the DP scheduling policy.

Lawrence, David↗

Machine and Application Aware Partitioning for Adaptive Mesh Refinement Applications. In: HPDC '17 Proceedings of the 26th International Symposium on High-Performance Parallel and Distributed Computing

Load balancing and partitioning are critical when it comes to parallel computations. Popular partitioning strategies based on space filling curves focus on equally dividing work. The partitions produced are independent of the architecture or the application. Given the ever-increasing relative cost of data movement and increasing heterogeneity of our architectures, it is no longer sufficient to only consider an equal partitioning of work. Minimizing communication costs are equally if not more important. Our hypothesis is that an unequal partitioning that minimizes communication costs significantly can scale and perform better than conventional equal-work partitioning schemes. This tradeoff is dependent on the architecture as well as the application. We validate our hypothesis in the context of a finite-element computation utilizing adaptive mesh-refinement. Our central contribution is a new partitioning scheme that minimizes the overall runtime of subsequent computations by performing architecture and application-aware non-uniform work assignment in order to decrease time to solution, primarily by minimizing data-movement. We evaluate our algorithm by comparing it against standard space-filling curve based partitioning algorithms and observing time-to-solution as well as energy-to-solution for solving Finite Element computations on adaptively refined meshes. We demonstrate excellent scalability of our new partition algorithm up to $262,144$ cores on ORNL's Titan and demonstrate that the proposed partitioning scheme reduces overall energy as well as time-to-solution for application codes by up to 22.0%.

Fernando, Milinda↗

Reduce Operations: Send Volume Balancing While Minimizing Latency

Communication hypergraph model was introduced in a two-phase setting for encapsulating multiple communication cost metrics (bandwidth and latency), which are proven to be important in parallelizing irregular applications. In the first phase, computational-task-to-processor assignment is performed with the objective of minimizing total volume while maintaining computational load balance. In the second phase, communication-task-to-processor assignment is performed with the objective of minimizing total number of messages while maintaining communication-volume balance. The reduce-communication hypergraph model suffers from failing to correctly encapsulate send-volume balancing. We propose a novel vertex weighting scheme that enables part weights to correctly encode send-volume loads of processors for send-volume balancing. The model also suffers from increasing the total communication volume during partitioning. To decrease this increase, we propose a method that utilizes the recursive bipartitioning framework and refines each bipartition by vertex swaps. For performance evaluation, we consider column-parallel SpMV, which is one of the most widely known applications in which the reduce-task assignment problem arises. Extensive experiments on 313 matrices show that, compared to the existing model, the proposed models achieve considerable improvements in all communication cost metrics. Furthermore, these improvements lead to an average decrease of 30% in parallel SpMV time on 512 processors for 70 matrices with high irregularity.

97 MATHEMATICS AND COMPUTING↗

Exploring utilization options of heterogeneous architectures for multi-physics simulations

Top supercomputers in the TOP500 list have transitioned from homogeneous node architectures toward heterogeneous manycore nodes with accelerators and CPUs. These new architectures present significant challenges to developers of large-scale multiphysics applications, especially at DOE laboratories that have invested heavily in scalable MPI codes over decades. Much of these scientific application porting efforts for the new heterogeneous architectures are focused on running the computation on the accelerators, which usually comprise > 90% of the FLOPS of the system, typically using one core per GPU to offload work. Furthermore, we describe an approach to utilizing the remaining FLOPS on a heterogeneous machine by running a portion of the computation on the CPUs cooperatively with the GPU computation. Orthogonally, we evaluate using the additional cores to also offload work to GPUs, effectively oversubscribing the GPUs. We present a proof-of-concept implementation in ARES, a multiphysics ALE-AMR code at LLNL. ARES uses a portability layer, RAJA, which enables us to utilize the same source code for both the CPU and the GPU. We develop an approach to utilize both types of processors cooperatively in a heterogeneous system. Our implementation divides the work between the computing resources via domain decomposition, and utilizes all cores of the CPU and all of the GPUs on the node for computation. Load balancing is necessary to use the heterogeneous resources effectively. We present preliminary results on early delivery pre-Sierra machines at LLNL, showing up to 40% performance benefit of oversubscribing the GPUs (with caveats), and up to an 18% performance benefit of using the CPUs on the heterogeneous nodes for computing in addition to using the GPUs.

97 MATHEMATICS AND COMPUTING↗

On Eliminating Synchronous Communication in Molecular Simulations to Improve Scalability

Molecular dynamics simulation, as a complementary tool to experimentation, has become an important methodology for the understanding and design of molecular systems as it provides access to properties that are difficult, impossible or prohibitively expensive to obtain experimentally. Many of the available software packages have been parallelized to take advantage of modern massively concurrent processing resources. The challenge in achieving parallel efficiency is commonly attributed to the fact that molecular dynamics algorithms are communication intensive. This paper illustrates how an appropriately chosen data distribution and asynchronous one-sided communication approach can be used to effectively deal with the data movement within the Global Arrays/ARMCI programming model framework. A new put_notify capability is presented here, allowing the implementation of the molecular dynamics algorithm without any explicit global or local synchronization or global data reduction operations. In addition, this push-data model is shown to very effectively allow hiding data communication behind computation. Rather than data movement or explicit global reductions, the implicit synchronization of the algorithm becomes the primary challenge for scalability. Without any explicit synchronous operations, the scalability of molecular simulations is shown to depend only on the ability to evenly balance computational load.

Straatsma, TP↗

Priority-enabled Load Balancing for Dispersed Computing

Opportunistic managed access to local in-network compute resources can improve the performance of distributed applications and reduce the dependence on shared network resources. Instead of backhauling application data to a centralized cloud data center for processing, networked services may be adaptively and continuously dispersed into shared compute resources that are closer to the source of need. While this approach has several benefits, support for mission-aware access to computation is often an afterthought, and is implemented as a brittle extension over traditional load-balancer solutions.In this work, we investigate the design of two priority-aware resource allocation strategies and two load-balancing dispatching strategies as first class citizens in an open-source dispersed computing middleware. We present a control theoretic analysis of these load-balancing primitives to identify weaknesses and strengths in our design, and recommend future directions. In parallel, we prototype two priority-aware allocation algorithms to validate our priority predictions. In initial experiments our prototype shows substantial gains in processing prioritized load. Finally, we make our source-code and experimental configurations open source.

Paulos, Aaron↗

Partitioning Models for Scaling Parallel Sparse Matrix-Matrix Multiplication

We investigate outer-product--parallel, inner-product--parallel, and row-by-row-product--parallel formulations of sparse matrix-matrix multiplication (SpGEMM) on distributed memory architectures. For each of these three formulations, we propose a hypergraph model and a bipartite graph model for distributing SpGEMM computations based on one-dimensional (1D) partitioning of input matrices. Here, we also propose a communication hypergraph model for each formulation for distributing communication operations. The computational graph and hypergraph models adopted in the first phase aim at minimizing the total message volume and balancing the computational loads of processors, whereas the communication hypergraph models adopted in the second phase aim at minimizing the total message count and balancing the message volume loads of processors. That is, the computational partitioning models reduce the bandwidth cost and the communication hypergraph models reduce the latency cost. Our extensive parallel experiments on up to 2048 processors for a wide range of realistic SpGEMM instances show that although the outer-product--parallel formulation scales better, the row-by-row-product--parallel formulation is more viable due to its significantly lower partitioning overhead and competitive scalability. For computational partitioning models, our experimental findings indicate that the proposed bipartite graph models are attractive alternatives to their hypergraph counterparts because of their lower partitioning overhead. Finally, we show that by reducing the latency cost besides the bandwidth cost through using the communication hypergraph models, the parallel SpGEMM time can be further improved up to 32%.

97 MATHEMATICS AND COMPUTING↗

Work stealing for GPU-accelerated parallel programs in a global address space framework

Task parallelism is an attractive approach to automatically load balance the computation in a parallel system and adapt to dynamism exhibited by parallel systems. Exploiting task parallelism through work stealing has been extensively studied in shared and distributed-memory contexts. In this paper, we study the design of a system that uses work stealing for dynamic load balancing of task-parallel programs executed on hybrid distributed-memory CPU-graphics processing unit (GPU) systems in a global-address space framework. We take into account the unique nature of the accelerator model employed by GPUs, the significant performance difference between GPU and CPU execution as a function of problem size, and the distinct CPU and GPU memory domains. We consider various alternatives in designing a distributed work stealing algorithm for CPU-GPU systems, while taking into account the impact of task distribution and data movement overheads. These strategies are evaluated using microbenchmarks that capture various execution configurations as well as the state-of-the-art CCSD(T) application module from the computational chemistry domain

97 MATHEMATICS AND COMPUTING↗

Profiling the BLAST bioinformatics application for load balancing on high-performance computing clusters

Abstract Background The Basic Local Alignment Search Tool (BLAST) is a suite of commonly used algorithms for identifying matches between biological sequences. The user supplies a database file and query file of sequences for BLAST to find identical sequences between the two. The typical millions of database and query sequences make BLAST computationally challenging but also well suited for parallelization on high-performance computing clusters. The efficacy of parallelization depends on the data partitioning, where the optimal data partitioning relies on an accurate performance model. In previous studies, a BLAST job was sped up by 27 times by partitioning the database and query among thousands of processor nodes. However, the optimality of the partitioning method was not studied. Unlike BLAST performance models proposed in the literature that usually have problem size and hardware configuration as the only variables, the execution time of a BLAST job is a function of database size, query size, and hardware capability. In this work, the nucleotide BLAST application BLASTN was profiled using three methods: shell-level profiling with the Unix “time” command, code-level profiling with the built-in “profiler” module, and system-level profiling with the Unix “gprof” program. The runtimes were measured for six node types, using six different database files and 15 query files, on a heterogeneous HPC cluster with 500+ nodes. The empirical measurement data were fitted with quadratic functions to develop performance models that were used to guide the data parallelization for BLASTN jobs. Results Profiling results showed that BLASTN contains more than 34,500 different functions, but a single function, RunMTBySplitDB, takes 99.12% of the total runtime. Among its 53 child functions, five core functions were identified to make up 92.12% of the overall BLASTN runtime. Based on the performance models, static load balancing algorithms can be applied to the BLASTN input data to minimize the runtime of the longest job on an HPC cluster. Four test cases being run on homogeneous and heterogeneous clusters were tested. Experiment results showed that the runtime can be reduced by 81% on a homogeneous cluster and by 20% on a heterogeneous cluster by re-distributing the workload. Discussion Optimal data partitioning can improve BLASTN’s overall runtime 5.4-fold in comparison with dividing the database and query into the same number of fragments. The proposed methodology can be used in the other applications in the BLAST+ suite or any other application as long as source code is available.

59 BASIC BIOLOGICAL SCIENCES↗

PYRANDA

Pyranda Is fortran 90 code which solves compact finite difference equations on a three dimensional fortran arrays, domain decomposed using the Message Passing Interface (MPI). The fortran source solves the linear system of equations which forms a pentadtagona1 matrix. Derivatives are used to construct hyperbolic partlal differential equation (PDE) operators, namely gradient, divergence, Laplacian, and curl operators. 5. Solution Method: The library provides a python Interface which facilitates problem setup and description. However, the core solver Is written in the fortran 90 language. The solver couples MPI communication and linear solves together so that a direct (non-iterative) solution to the linear system of equations can be computed with minimal communication. Domain decomposition is limited to a uniform data size but provides perfect load balancing. 6. Computer for which the code Is written:

Olson, BrittonJ.↗

A Novel Coarsening Method for Scalable and Efficient Mesh Generation

In this paper, we propose a novel mesh coarsening method called brick coarsening method. The proposed method can be used in conjunction with any graph partitioners and scales to very large meshes. This method reduces problem space by decomposing the original mesh into fixed-size blocks of nodes called bricks, layered in a similar way to conventional brick laying, and then assigning each node of the original mesh to appropriate brick. Our experiments indicate that the proposed method scales to very large meshes while allowing simple RCB partitioner to produce higher-quality partitions with significantly less edge cuts. Our results further indicate that the proposed brick-coarsening method allows more complicated partitioners like PT-Scotch to scale to very large problem size while still maintaining good partitioning performance with relatively good edge-cut metric. Graph partitioning is an important problem that has many scientific and engineering applications in such areas as VLSI design, scientific computing, and resource management. Given a graph G = (V,E), where V is the set of vertices and E is the set of edges, (k-way) graph partitioning problem is to partition the vertices of the graph (V) into k disjoint groups such that each group contains roughly equal number of vertices and the number of edges connecting vertices in different groups is minimized. Graph partitioning plays a key role in large scientific computing, especially in mesh-based computations, as it is used as a tool to minimize the volume of communication and to ensure well-balanced load across computing nodes. The impact of graph partitioning on the reduction of communication can be easily seen, for example, in different iterative methods to solve a sparse system of linear equation. Here, a graph partitioning technique is applied to the matrix, which is basically a graph in which each edge is a non-zero entry in the matrix, to allocate groups of vertices to processors in such a way that many of matrix-vector multiplication can be performed locally on each processor and hence to minimize communication. Furthermore, a good graph partitioning scheme ensures the equal amount of computation performed on each processor. Graph partitioning is a well known NP-complete problem, and thus the most commonly used graph partitioning algorithms employ some forms of heuristics. These algorithms vary in terms of their complexity, partition generation time, and the quality of partitions, and they tend to trade off these factors. A significant challenge we are currently facing at the Lawrence Livermore National Laboratory is how to partition very large meshes on massive-size distributed memory machines like IBM BlueGene/P, where scalability becomes a big issue. For example, we have found that the ParMetis, a very popular graph partitioning tool, can only scale to 16K processors. An ideal graph partitioning method on such an environment should be fast and scale to very large meshes, while producing high quality partitions. This is an extremely challenging task, as to scale to that level, the partitioning algorithm should be simple and be able to produce partitions that minimize inter-processor communications and balance the load imposed on the processors. Our goals in this work are two-fold: (1) To develop a new scalable graph partitioning method with good load balancing and communication reduction capability. (2) To study the performance of the proposed partitioning method on very large parallel machines using actual data sets and compare the performance to that of existing methods. The proposed method achieves the desired scalability by reducing the mesh size. For this, it coarsens an input mesh into a smaller size mesh by coalescing the vertices and edges of the original mesh into a set of mega-vertices and mega-edges. A new coarsening method called brick algorithm is developed in this research. In the brick algorithm, the zones in a given mesh are first grouped into fixed size blocks called bricks. These brick are then laid in a way similar to conventional brick laying technique, which reduces the number of neighboring blocks each block needs to communicate. Contributions of this research are as follows: (1) We have developed a novel method that scales to a really large problem size while producing high quality mesh partitions; (2) We measured the performance and scalability of the proposed method on a machine of massive size using a set of actual large complex data sets, where we have scaled to a mesh with 110 million zones using our method. To the best of our knowledge, this is the largest complex mesh that a partitioning method is successfully applied to; and (3) We have shown that proposed method can reduce the number of edge cuts by as much as 65%.

97 MATHEMATICS AND COMPUTING↗

ESnet/JLab FPGA Accelerated Transport

To increase the science rate for high data rates/volumes, Thomas Jefferson National Accelerator Facility (JLab) has partnered with Energy Sciences Network (ESnet) to define an edge to data center traffic shaping / steering transport capability featuring data event aware network shaping and forwarding. The keystone of this ESnet+JLab FPGA Accelerated Transport (EJFAT) is the joint development of an AI/ML directed dynamic compute work Load Balancer (LB) of UDP streamed data. The LB is a suite consisting of a Field Programmable Gate Array (FPGA) executing the dynamically configurable, low fixed latency LB data plane featuring real-time packet redirection and high throughput, and a control plane running on the FPGA host computer that monitors network and compute farm telemetry in order to make dynamic AI/ML guided decisions for destination compute host redirection/load balancing and destination resource provisioning. The LB provides for three-tier horizontal scaling across LB suites, core compute hosts, and CPUs within a host. The LB effectively provides seamless integration of edge/core computing to support direct experimental data processing for immediate use by JLab science programs and others such as the EIC as well as data centers of the future requiring high throughput and low latency for both hot and cooled data for both running experiment data acquisition systems and data center use cases.

97 MATHEMATICS AND COMPUTING↗

GraphBLAST: A High-Performance Linear Algebra-based Graph Framework on the GPU

High-performance implementations of graph algorithms are challenging to implement on new parallel hardware such as GPUs, because of three challenges: (1) difficulty of coming up with graph building blocks, (2) load imbalance on parallel hardware, and (3) graph problems having low arithmetic intensity. To address these challenges, GraphBLAS is an innovative, on-going effort by the graph analytics community to propose building blocks based in sparse linear algebra, which will allow graph algorithms to be expressed in a performant, succinct, composable and portable manner. In this paper, we examine the performance challenges of a linear algebra-based approach to building graph frameworks and describe new design principles for overcoming these bottlenecks. Among the new design principles is exploiting input sparsity, which allows users to write graph algorithms without specifying push and pull direction. Exploiting output sparsity allows users to tell the backend which values of the output in a single vectorized computation they do not want computed. Load-balancing is an important feature for balancing work amongst parallel workers. We describe the important load-balancing features for handling graphs with different characteristics. The design principles described in this paper have been implemented in "GraphBLAST", the first open-source linear algebra-based graph framework on GPU targeting high-performance computing. The results show that on a single GPU, GraphBLAST has on average at least an order of magnitude speedup over previous GraphBLAS implementations SuiteSparse and GBTL, comparable performance to the fastest GPU hardwired primitives and shared-memory graph frameworks Ligra and Gunrock, and better performance than any other GPU graph framework, while offering a simpler and more concise programming model.

97 MATHEMATICS AND COMPUTING↗

An Efficient Storage-Driven Machine Learning Model for Performance in the Era of Multimodal Scientific Data

Scientific workflows are increasingly relying on machine learning (ML), simulation, and hybrid techniques to predict, understand, and optimize the behavior of complex experiments. High-performance computing has greatly improved researchers’ ability to acquire diverse data modalities in these workflows. Recent studies suggest that the performance of machine learning models can be improved by integrating data from various sources. Unfortunately, these workloads pose unprecedent pressure on the network storage to meet the demands associated with accessing these multimodal data. To mitigate the impact of intensive IO, we propose a solution that utilizes a multi-tier High-Performance Computing (HPC) distributed storage and data processing framework, placing computation where the data resides for better performance. By adopting this project, the scientific community will gain new opportunities to explore multimodal storage-driven possibilities, integrating multiple scientific data sources with advanced streaming frameworks. Additionally, our framework effectively utilizes computing resources and bridges the gaps identified by HPC experts. Our proposed approach tackles scalability and persistence challenges by leveraging native persistency, which has posed difficulties in traditional approaches. Furthermore, we seek to enhance fault-tolerance and load-balance of computations by leveraging real-time streaming in diverse scientific computing environments, thereby propelling advanced scientific computing research into the next generation.

97 MATHEMATICS AND COMPUTING↗

High Performance Molecular Dynamic Simulation on Single and Multi-GPU Systems

The programming techniques supported and employed on these GPUs and Multi-GPUs systems are not sufficient to address problems exhibiting irregular, and unbalanced workload such as Molecular Dynamic (MD) simulations of systems with non-uniform densities. In this paper, we propose a task-based dynamic load-balancing solution to employ on MD simulations for single- and multi-GPU systems. The solution allows load balancing at a finer granularity than what is supported in existing APIs such as NVIDIA’s CUDA. Experimental results with a single-GPU configuration show that our fine-grained task solution can utilize the hardware more efficiently than the CUDA scheduler. On multi-GPU systems, our solution achieves near-linear speedup, load balance, and significant performance improvement over techniques based on standard CUDA APIs.

97 MATHEMATICS AND COMPUTING↗