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 685 records · Page 38

Utility functions and resource management in an oversubscribed heterogeneous computing environment

We model an oversubscribed heterogeneous computing system where tasks arrive dynamically and a scheduler maps the tasks to machines for execution. The environment and workloads are based on those being investigated by the Extreme Scale Systems Center at Oak Ridge National Laboratory. Utility functions that are designed based on specifications from the system owner and users are used to create a metric for the performance of resource allocation heuristics. Each task has a time-varying utility (importance) that the enterprise will earn based on when the task successfully completes execution. We design multiple heuristics, which include a technique to drop low utility-earning tasks, to maximize the total utility that can be earned by completing tasks. The heuristics are evaluated using simulation experiments with two levels of oversubscription. The results show the benefit of having fast heuristics that account for the importance of a task and the heterogeneity of the environment when making allocation decisions in an oversubscribed environment. Furthermore, the ability to drop low utility-earning tasks allow the heuristics to tolerate the high oversubscription as well as earn significant utility.

97 MATHEMATICS AND COMPUTING↗

PANTHER: A Programmable Architecture for Neural Network Training Harnessing Energy-Efficient ReRAM

The wide adoption of deep neural networks has been accompanied by ever-increasing energy and performance demands due to the expensive nature of training them. Additionally, numerous special-purpose architectures have been proposed to accelerate training: both digital and hybrid digital-analog using resistive RAM (ReRAM) crossbars. ReRAM-based accelerators have demonstrated the effectiveness of ReRAM crossbars at performing matrix-vector multiplication operations that are prevalent in training. However, they still suffer from inefficiency due to the use of serial reads and writes for performing the weight gradient and update step. A few works have demonstrated the possibility of performing outer products in crossbars, which can be used to realize the weight gradient and update step without the use of serial reads and writes. However, these works have been limited to low precision operations which are not sufficient for typical training workloads. Moreover, they have been confined to a limited set of training algorithms for fully-connected layers only. To address these limitations, we propose a bit-slicing technique for enhancing the precision of ReRAM-based outer products, which is substantially different from bit-slicing for matrix-vector multiplication only. We incorporate this technique into a crossbar architecture with three variants catered to different training algorithms. To evaluate our design on different types of layers in neural networks (fully-connected, convolutional, etc.) and training algorithms, we develop PANTHER, an ISA-programmable training accelerator with compiler support. Our design can also be integrated into other accelerators in the literature to enhance their efficiency. Our evaluation shows that PANTHER achieves up to 8.02×, 54.21×, and 103× energy reductions as well as 7.16×, 4.02×, and 16× execution time reductions compared to digital accelerators, ReRAM-based accelerators, and GPUs, respectively.

42 ENGINEERING↗

Elastic Resource Management for Deep Learning Applications in a Container Cluster

The increasing demand for learning from massive datasets is restructuring our economy. Effective learning, however, involves nontrivial computing resources. Most businesses utilize commercial infrastructure providers (e.g., AWS) to host their computing clusters in the cloud, where various jobs compete for available resources. While cloud resource management is a fruitful research field that has made many advances in production, such as Kubernetes and YARN, few efforts have been invested to further optimize the system performance, especially for deep learning (DL) training jobs in a container cluster. This work introduces FlowCon, a system that is able to monitor the individual evaluation functions of DL jobs at runtime, and thus to make placement decisions on resource allocations elastically. Here, we present a detailed design and implementation of FlowCon and conduct intensive experiments over various DL models. The results demonstrate that FlowCon significantly improves DL job completion time and resource utilization efficiency, compared to default systems. According to the results, FlowCon is able to improve the completion time by up to 68.8% and meanwhile, reduce the makespan by 18.0%, in the presence of various DL job workloads.

97 MATHEMATICS AND COMPUTING↗

Cooling Matters: Benchmarking Large Language Models and Vision-Language Models on Liquid-Cooled Versus Air-Cooled H100 GPU Systems

The unprecedented growth in artificial intelligence (AI) workloads, recently dominated by large language models (LLMs) and vision-language models (VLMs), has intensified power and cooling demands in data centers. This study benchmarks LLMs and VLMs on two HGX nodes, each with 8× NVIDIA H100 graphics processing units (GPUs), using liquid and air cooling. Leveraging GPU Burn, Weights & Biases, and IPMItool, we collect detailed thermal, power, and computation data. Results show that the liquid-cooled systems maintain GPU temperatures between 41-50$^\circ$C, while the air-cooled counterparts fluctuate between 54-72$^\circ$C under load. This thermal stability of liquid-cooled systems yields 17% higher performance (54 TFLOPs/ GPU vs. 46 TFLOPs/GPU), performance-per-watt, reduced energy overhead, and greater system efficiency than the air-cooled counterparts. These findings underscore the energy and sustainability benefits of liquid cooling, offering a compelling path forward for hyperscale data centers seeking to optimize AI infrastructure. https://github.com/iscaas/Cooling-Matters.

Latif, Imran↗

A performance model for GPUs with caches

To exploit the abundant computational power of the world's fastest supercomputers, an even workload distribution to the typically heterogeneous compute devices is necessary. While relatively accurate performance models exist for conventional CPUs, accurate performance estimation models for modern GPUs do not exist. This paper presents two accurate models for modern GPUs: a sampling-based linear model, and a model based on machine-learning (ML) techniques which improves the accuracy of the linear model and is applicable to modern GPUs with and without caches. We first construct the sampling-based linear model to predict the runtime of an arbitrary OpenCL kernel. Based on an analysis of NVIDIA GPUs' scheduling policies we determine the earliest sampling points that allow an accurate estimation. The linear model cannot capture well the significant effects that memory coalescing or caching as implemented in modern GPUs have on performance. We therefore propose a model based on ML techniques that takes several compiler-generated statistics about the kernel as well as the GPU's hardware performance counters as additional inputs to obtain a more accurate runtime performance estimation for modern GPUs. We demonstrate the effectiveness and broad applicability of the model by applying it to three different NVIDIA GPU architectures and one AMD GPU architecture. On an extensive set of OpenCL benchmarks, on average, the proposed model estimates the runtime performance with less than 7 percent error for a second-generation GTX 280 with no on-chip caches and less than 5 percent for the Fermi-based GTX 580 with hardware caches. On the Kepler-based GTX 680, the linear model has an error of less than 10 percent. On an AMD GPU architecture, Radeon HD 6970, the model estimates with 8 percent of error rates. As a result, the proposed technique outperforms existing models by a factor of 5 to 6 in terms of accuracy.

97 MATHEMATICS AND COMPUTING↗

CoreTSAR: Core Task-Size Adapting Runtime

Heterogeneity continues to increase at all levels of computing, with the rise of accelerators such as GPUs, FPGAs, and other co-processors into everything from desktops to supercomputers. As a consequence, efficiently managing such disparate resources has become increasingly complex. CoreTSAR seeks to reduce this complexity by adaptively worksharing parallel-loop regions across compute resources without requiring any transformation of the code within the loop. Lastly, our results show performance improvements of up to three-fold over a current state-of-the-art heterogeneous task scheduler as well as linear performance scaling from a single GPU to four GPUs for many codes. In addition, CoreTSAR demonstrates a robust ability to adapt to both a variety of workloads and underlying system configurations.

97 MATHEMATICS AND COMPUTING↗

Implementation and Tuning of Batched Cholesky Factorization and Solve for NVIDIA GPUs

Many problems in engineering and scientific computing require the solution of a large number of small systems of linear equations. Due to their high processing power, Graphics Processing Units became an attractive target for this class of problems, and routines based on the LU and the QR factorization have been provided by NVIDIA in the cuBLAS library. This work addresses the situation where the systems of equations are symmetric positive definite. The paper describes the implementation and tuning of the kernels for the Cholesky factorization and the forward and backward substitution. Targeted workloads involve the solution of thousands of linear systems of the same size, where the focus is on matrix dimensions from 5 by 5 to 100 by 100. Due to the lack of a cuBLAS Cholesky factorization, execution rates of cuBLAS LU and cuBLAS QR are used for comparison against the proposed Cholesky factorization in this work. Execution rates of forward and backward substitution routines are compared to equivalent cuBLAS routines. Comparisons against optimized multicore implementations are also presented. Superior performance is reached in all cases.

Computer Science↗

ACStor: Optimizing Access Performance of Virtual Disk Images in Clouds

In virtualized data centers, virtual disk images (VDIs) serve as the containers in virtual environment, so their access performance is critical for the overall system performance. Some distributed VDI chunk storage systems have been proposed in order to alleviate the I/O bottleneck for VM management. As the system scales up to a large number of running VMs, however, the overall network traffic would become unbalanced with hot spots on some VMs inevitably, leading to I/O performance degradation when accessing the VMs. Here, we propose an adaptive and collaborative VDI storage system (ACStor) to resolve the above performance issue. In comparison with the existing research, our solution is able to dynamically balance the traffic workloads in accessing VDI chunks, based on the run-time network state. Specifically, compute nodes with lightly loaded traffic will be adaptively assigned more chunk access requests from remote VMs and vice versa, which can effectively eliminate the above problem and thus improves the I/O performance of VMs. We also implement a prototype based on our ACStor design, and evaluate it by various benchmarks on a real cluster with 32 nodes and a simulated platform with 256 nodes. Experiments show that under different network traffic patterns of data centers, our solution achieves up to 2-8 performance gain on VM booting time and VM’s I/O throughput, in comparison with the other state-of-the-art approaches.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗

Unraveling Network-induced Memory Contention: Deeper Insights with Machine Learning

Remote Direct Memory Access (RDMA) is expected to be an integral communication mechanism for future exascale systems enabling asynchronous data transfers, so that applications may fully utilize CPU resources while simultaneously sharing data amongst remote nodes. We examine Network-induced Memory Contention (NiMC) on Infiniband networks. We expose the interactions between RDMA, main-memory and cache, when applications and out-of-band services compete for memory resources. We then explore NiMCs resulting impact on application-level performance. For a range of hardware technologies and HPC workloads, we quantify NiMC and show that NiMCs impact grows with scale resulting in up to 3X performance degradation at scales as small as 8K processes even in applications that previously have been shown to be performance resilient in the presence of noise. In addition, this work examines the problem of predicting NiMC's impact on applications by leveraging machine learning and easily accessible performance counters. This approach provides additional insights about the root cause of NiMC and facilitates dynamic selection of potential solutions. Finally, we evaluated three potential techniques to reduce NiMCs impact, namely hardware offloading, core reservation and network throttling.

97 MATHEMATICS AND COMPUTING↗

Large-Scale and Extreme-Scale Computing with Stranded Green Power: Opportunities and Costs

Power consumption and associated carbon emissions are increasingly critical challenges for large-scale computing. Recent research proposes exploiting stranded power-uneconomic renewable power-for green supercomputing in a system called Zero-Carbon Cloud (ZCCloud). These efforts studied production supercomputing workloads on stranded-power based computing resources, demonstrating their achievable productivity. We explore economic viability of stranded-power based supercomputing, using three datacenter total-cost-of-ownership (TOO) models to study cost-effectiveness. These studies show that ZCCloud's approach can be cost-effective in the USA today, and is even more attractive in regions with higher power prices (e.g., Japan, Germany), achieving cost advantages as large as 50 percent. Environmental and power-grid benefits are a further advantage. We also explore the sensitivity of these results to changes in hardware TOO; cheaper hardware or longer lifetimes magnify the attractiveness of stranded-power based approaches, yielding advantages as large as 91 percent. These results are robust across different TCO models. Lastly, we study extreme-scale supercomputers ( >100 MW), finding stranded-power can increase peak capability per cost by as much as 80 percent.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Dynamic Adaptable Asynchronous Progress Model for MPI RMA Multiphase Applications

Casper is a process-based asynchronous progress model for MPI one-sided communication on multi-and many-core architectures. The one-sided communication is not truly one-sided in most MPI implementations: the target process still relies on software progress to complete incoming operations. Casper allows the user to specify an arbitrary number of cores dedicated to background ghost processes and transparently redirects the RMA operations to ghost processes by utilizing the PMPI redirection and MPI-3 shared-memory technologies. Although Casper benefits applications that suffer from lack of asynchronous progress, the operation redirection design might not support complex multiphase applications effectively, which often involve dynamically changing communication density and computing workloads. In this paper, we present an adaptive mechanism in Casper to address the limitation of static asynchronous progress in multiphase applications. For this we exploit two adaptive strategies, a user-guided strategy and a fully transparent and automatic strategy based on self-profiling and prediction, to dynamically reconfigure the asynchronous progress in Casper according to real-time performance characteristics during multiphase execution. We evaluate the adaptive approaches in both microbenchmarks and a real quantum chemistry application suite, NWChem, on the Cray XC30 supercomputer and an Intel Omni-Path cluster.

97 MATHEMATICS AND COMPUTING↗

Persistent Octrees for Parallel Mesh Refinement Through Non-Volatile Byte-Addressable Memory

We report that octree-based mesh adaptation has enabled simulations of complex physical phenomena. Existing meshing algorithms were proposed with the assumption that computer memory is volatile. Consequently, for failure recovery, the in-core algorithms need to save memory states as snapshots with slow file I/Os. The out-of-core algorithms store octants on disks for persistence. However, neither of them was designed to leverage unique characteristics of non-volatile byte-addressable memory (NVBM). In this paper, we propose a novel data structure Distributed Persistent Merged octree (DPM-octree) for both meshing and in-memory storage of persistent octrees using NVBM. It is a multi-version data structure and can recover from failures using its earlier persistent version stored in NVBM. In addition, we design a feature-directed sampling approach to help dynamically transform the DPM-octree layout for reducing NVBM-induced memory write latency. DPM-octree uses parity trees which are created using erasure coding and stored in NVBM to support low-latency in-memory octant recovery after data loss. DPM-octree has been successfully integrated with Gerris software for simulation of fluid dynamics. Finally, our experimental results with real-world scientific workloads show that DPM-octree scales up to 1.1 billion mesh elements with 1000 processors on the Titan supercomputer.

97 MATHEMATICS AND COMPUTING↗

Harnessing Data Movement in Virtual Clusters for In-Situ Execution

As a result of increasing data volume and velocity, Big Data science at exascale has shifted towards the in-situ paradigm, where large scale simulations run concurrently alongside data analytics. With in-situ, data generated from simulations can be processed while still in memory, thereby avoiding the slow storage bottleneck. However, running simulations and analytics together on shared resources will likely result in substantial contention if left unmanaged, as demonstrated in this work, leading to much reduced efficiency of simulations and analytics. Recently, virtualization technologies such as Linux containers have been widely applied to data centers and physical clusters to provide highly efficient and elastic resource provisioning for consolidated workloads including scientific simulations and data analytics. In this paper, we investigate to facilitate network traffic manipulation and reduce mutual interference on the network for in-situ applications in virtual clusters. In order to dynamically allocate the network bandwidth when it is needed, we adopt SARIMA-based techniques to analyze and predict MPI traffic issued from simulations. Although this can be an effective technique, the naïve usage of network virtualization can lead to performance degradation for bursty asynchronous transmissions within an MPI job. Here, we analyze and resolve this performance degradation in virtual clusters.

97 MATHEMATICS AND COMPUTING↗

Managing Rich Metadata in High-Performance Computing Systems Using a Graph Model

High-performance computing (HPC) systems generate huge amounts of metadata about different entities such as jobs, users, and files. Existing systems can efficiently record and manage part of these metadata, mainly the POSIX metadata of data files (e.g., file size, name, and permissions mode). But another important set of metadata, referred to as "rich" metadata in this study, which record not only wider range of entities (e.g., running processes and jobs) but also more complex relationships between them, are mostly missing in current HPC systems. Yet such rich metadata are critical for supporting many advanced data management functions such as identifying data sources and parameters behind a given result; auditing data usage; or understanding details about how inputs are transformed into outputs. To uniformly and efficiently manage the rich metadata generated in HPC systems, We propose to utilize a graph model in this study. We identify the key challenges of implementing such a graph-based HPC rich metadata management system and present GraphMeta, a graph-based rich metadata management system designed and optimized for HPC platforms, to tackle these challenges. In conclusion, extensive evaluations on both synthetic and real HPC metadata workloads show its advantages in both performance and scalability compared with existing solutions.

97 MATHEMATICS AND COMPUTING↗

Customizable Scale-Out Key-Value Stores

Enterprise KV stores are often not well suited for HPC applications, and thus cumbersome end-to-end KV design customization is required to meet the needs of modern HPC applications. To this end, in this article we present bespoKV, an adaptive, extensible, and scale-out KV store framework. bespoKV decouples the KV store design into the control plane for distributed management and the data plane for local data store. For the control plane, bespoKVprovides pre-built modules, called controlets, supporting common distributed functionalities (e.g., replication, consistency, and topology) and their various combinations. This decoupling allows bespoKV to take a user-provided single-server KV store, called a datalet, and transparently enables a scalable and fault-tolerant distributed KV store service. The resulting distributed stores are also adaptive to consistency or topology requirement changes and can be easily extended for new types of services. Such specializations enable innovative uses of KV stores in HPC applications, especially for emerging applications that utilize KV-friendly workloads. We evaluate bespoKV in a local testbed as well as in a public cloud settings. Experiments show that bespoKV-enabled distributed KV stores scale horizontally to a large number of nodes, and performs comparably and sometimes 1.2× to 2.6× better than the state-of-the-art systems.

97 MATHEMATICS AND COMPUTING↗

Enabling Scalable and Extensible Memory-mapped Datastores in Userspace

Exascale workloads are expected to incorporate data-intensive processing in close coordination with traditional physics simulations. These emerging scientific, data-analytics and machine learning applications need to access a wide variety of datastores in flat files and structured databases. Programmer productivity is greatly enhanced by mapping datastores into the application process's virtual memory space to provide a unified “in-memory” interface. Currently, memory mapping is provided by system software primarily designed for generality and reliability. However, scalability at high concurrency is a formidable challenge on exascale systems. Also, there is a need for extensibility to support new datastores potentially requiring HPC data transfer services. In this article, we present UMap , a scalable and extensible userspace service for memory-mapping datastores. Furthermore, through decoupled queue management, concurrency aware adaptation, and dynamic load balancing, UMap enables application performance to scale even at high concurrency. We evaluate UMap in data-intensive applications, including sorting, graph traversal, database operations, and metagenomic analytics. Our results show that UMap as a userspace service outperforms an optimized kernel-based service across a wide range of intra-node concurrency by 1.22-1.9 × . We performed two case studies to demonstrate UMap 's extensibility. First, a new datastore residing in remote memory is incorporated into UMap as an application-specific plugin. Second, we present a persistent memory allocator Metall built atop UMap for unified storage/memory.

97 MATHEMATICS AND COMPUTING↗

GSoFa: Scalable Sparse Symbolic LU Factorization on GPUs

Decomposing a matrix $\mathbf {A}$ into a lower matrix $\mathbf {L}$ and an upper matrix $\mathbf {U}$, which is also known as LU decomposition, is an essential operation in numerical linear algebra. For a sparse matrix, LU decomposition often introduces more nonzero entries in the $\mathbf {L}$ and $\mathbf {U}$ factors than in the original matrix. A symbolic factorization step is needed to identify the nonzero structures of $\mathbf {L}$ and $\mathbf {U}$ matrices. Attracted by the enormous potentials of the Graphics Processing Units (GPUs), an array of efforts have surged to deploy various LU factorization steps except for the symbolic factorization, to the best of our knowledge, on GPUs. This article introduces gSoFa, the first GPU-based symbolic factorization design with the following three optimizations to enable scalable LU symbolic factorization for nonsymmetric pattern sparse matrices on GPUs. First, here we introduce a novel fine-grained parallel symbolic factorization algorithm that is well suited for the Single Instruction Multiple Thread (SIMT) architecture of GPUs. Second, we tailor supernode detection into a SIMT friendly process and strive to balance the workload, minimize the communication and saturate the GPU computing resources during supernode detection. Third, we introduce a three-pronged optimization to reduce the excessive space consumption problem faced by multi-source concurrent symbolic factorization. Taken together, gSoFa achieves up to 31× speedup from 1 to 44 Summit nodes (6 to 264 GPUs) and outperforms the state-of-the-art CPU project, on average, by 5×. Notably, gSoFa also achieves up to 47 percent of the peak memory throughput of a V100 GPU in the Summit Supercomputer.

97 MATHEMATICS AND COMPUTING↗

IRIS: A Performance-Portable Framework for Cross-Platform Heterogeneous Computing

From edge to exascale, computer architectures are becoming more heterogeneous and complex. The systems typically have fat nodes, with multicore CPUs and multiple hardware accelerators such as GPUs, FPGAs, and DSPs. This complexity is causing a crisis in programming systems and performance portability. Several programming systems are working to address these challenges, but the increasing architectural diversity is forcing software stacks and applications to be specialized for each architecture. As we show, all of these approaches critically depend on their software framework for discovery, execution, scheduling, and data orchestration. To address this challenge, we believe that a more agile and proactive software framework is essential to increase performance portability and improve user productivity. To this end, we have designed and implemented IRIS: a performance-portable framework for cross-platform heterogeneous computing. IRIS can discover available resources, manage multiple diverse programming platforms (e.g., CUDA, Hexagon, HIP, Level Zero, OpenCL, OpenMP) simultaneously in the same execution, respect data dependencies, orchestrate data movement proactively, and provide for user-configurable scheduling. To simplify data movement, IRIS introduces a shared virtual device memory with relaxed consistency among different heterogeneous devices. IRIS also adds an automatic kernel workload partitioning technique using the polyhedral model so that it can resize kernels for a wide range of devices. Our evaluation on three architectures, ranging from Qualcomm Snapdragon to a Summit supercomputer node, shows that IRIS improves portability across a wide range of diverse heterogeneous architectures with negligible overhead.

97 MATHEMATICS AND COMPUTING↗