Search

In this article, a scalable autonomous separation assurance framework is proposed for high-density en route airspace sectors with heterogeneous aircraft objectives. To handle the complex dynamic decision making under uncertainty, multi-agent reinforcement learning is used in a decentralized approach with each aircraft being represented as an agent. Based on this, each agent locally solves the separation assurance problem, allowing the framework to scale to a large number of aircraft. In addition, each agent has the ability to learn the intention of the intruder aircraft, which is essential in environments with heterogeneous agents. Numerical experiments are performed in a real-time air traffic simulator. The results demonstrate that the proposed framework is able to effectively ensure the safe separation of heterogeneous agents, while also optimizing the intrinsic agent objectives in high-density en route airspace sectors. In addition, the efficiency of the proposed framework is demonstrated and shown to provide real-time decision making for separation assurance. Note to Practitioners—In commercial aviation, the workload of human air traffic controllers increases with the growth of air traffic density. Robust decision making systems to augment human air traffic controllers allows for increased air traffic without increased workload, resulting in a safer airspace environment. In addition, advanced air mobility (AAM) is concerned with low-altitude airspace operations with both human and autonomous pilots. In this environment, autonomous real-time separation assurance systems are required. Most traditional separation assurance approaches fail to handle stochastic and high-density environments, rendering them inapplicable to future high-density traditional airspace and the envisioned low-altitude AAM operations. Therefore, it is important to study decentralized approaches that can place the separation assurance problem as an intrinsic objective of each aircraft to ensure cooperation in high-density airspace. With this consideration, a scalable, decentralized autonomous separation assurance framework capable of handling heterogeneous agents is proposed in this article. This framework is able to perceive the current air traffic environment and select speed advisories to ensure safe separation requirements, while balancing intrinsic objectives such as minimizing delay. While one limitation of multi-agent reinforcement learning is the long training time, this article demonstrates how the framework also can leverage modern computing clusters to significantly reduce training time without sacrificing performance. [ABSTRACT FROM AUTHOR]

This report describes The University of Texas Student Cluster Competition team’s effort to reproduce the results of “MemXCT: memory-centric X-ray CT reconstruction with massive parallelization” (Hidayetoğlu et al., 2019). The article details a new memory-centric approach that reconstructs X-ray computed tomography (XCT) from noisy raw data. In our reproduction experiments, we utilized Microsoft Azure’s CycleCloud tool to provision, orchestrate, and manage our computing cluster in the cloud. In particular, we scheduled and benchmarked reconstruction workloads using Azure’s CPU-based HC44rs and GPU-based NC12s v2 virtual machine (VM) types to evaluate the scalability properties of the reconstruction approach and the performance differences between architectures. The HC44rs VMs contained 44 Intel Xeon Platinum cores, while the NC12s v2 VM was equipped with two NVIDIA P100 GPUs. We used a recent version of Intel’s compiler stack with the MKL library for our CPU code along with CUDA 11.1 on GPUs. Overall, our results confirm the findings of the original article, demonstrating similar acceleration on GPUs and scalability properties on CPUs. Digital artifacts from these experiments are available at: 10.5281/zenodo.5598108 [ABSTRACT FROM AUTHOR]

Recently, with the large-scale outbreak of the global financial crisis and public safety incidents (such as COVID-19), high-performance computing has been widely applied to risk prediction, vaccine development, and other fields. In scenarios where high-performance computing infrastructure responds to the instantaneous explosion of computing demands, a crucial issue is to provide large-scale flexible allocation and adjustment of computing capability by rapidly constructing computing clusters. Existing large-scale computing cluster deployment solutions usually utilize source code deployment or other deployment tools. The great challenge of existing deployment methods is to reduce excessive image distribution time and refrain from configuration defects. In this article, we design an intelligent distributed registry deployment (IDRD) architecture based on the OpenStack cloud platform, which adaptively places distributed image repositories using the containerized deployment of multiple registries. We propose a server load priority algorithm to solve multiple registries placement problems in IDRD. Furthermore, we devise a clustering algorithm based on demand density that can optimize the global performance of IDRD and improve large-scale cluster load balancing capabilities, which has been implemented in the TianHe Supercomputing environment. Extensive experimental results demonstrate that IDRD can effectively reduce 30% − 50% of the distribution time of component images and significantly improve the efficiency of large-scale cluster deployment. [ABSTRACT FROM AUTHOR]

For the rapid growth computation requirements in big data and artificial intelligence area, CPU-GPU heterogeneous clusters can provide more powerful computing capacity compared to CPU clusters. The high parallel computing capabilities of GPUs greatly accelerate computation-intensive applications. And the number of GPUs on single computing node is scalable, which greatly improves the computing capacity of the cluster under the condition of limited cluster size. However, there is a lack of the effective load-balancing scheduling model in multi-GPU hardware environment. This article proposes AEML, an acceleration engine for multi-GPU load-balancing in distributed heterogeneous environment. AEML can effectively integrate GPUs into distributed processing framework and achieve great load-balance among multiple heterogeneous GPUs. We propose a heterogeneous task execution model based on multiple GPUs and multiple streams (MGMS), which can effectively balance the workload of multiple GPUs. MGMS model utilizes four core techniques: a fine-grained task mapping mechanism, a device resource unified management scheme, a novel resource-aware GPU task scheduling strategy, and a feedback-based streams adjustment scheme. The implementation of AEML system is based on Spark 3.0.0 and NVIDIA CUDA 10.0. We comprehensively evaluate the performance of AEML with multiple typical benchmarks. Experimental results show that AEML can fully exploit the computing power of GPUs and achieve great load-balance among multiple heterogeneous GPUs. [ABSTRACT FROM AUTHOR]

In this communication, a novel message passing interface (MPI) parallel algorithm for nodal discontinuous Galerkin time-domain (NDGTD) method has been developed. A unified MPI + MPI technique has been introduced for extreme parallelism on a large-scale computer cluster. Through the data transmission between CPU nodes using MPI persistent nonblocking two-side communication and the direct data connection between processors in the same node via MPI shared memory windows, a two-layered parallel architecture is implemented to minimize the communication. To further accelerate the solution of the multiscale problems, the local time stepping (LTS) technique has been employed in the NDGTD method. A fast time step estimation method has been presented in this communication. With high overlap between the information transmission and the data calculation, the proposed MPI + MPI scheme overcomes the degradation of the parallel efficiency of the pure MPI technique in the scenario of the LTS technique and the large-scale CPU cores. Up to 94% parallel efficiency in 6400 CPU cores is achieved for the average single-core loading about 1700 finite elements, and 18 times acceleration for time step estimation can be obtained with the fourth-order basis function. Three practical complex examples are given to demonstrate a good performance of the proposed method. [ABSTRACT FROM AUTHOR]

Big data frameworks such as Apache Spark is becoming prominent to perform large-scale data analytics jobs in various domains. However, due to limited resource availability, the local or on-premise computing resources are often not sufficient to run these jobs. Therefore, public cloud resources can be hired on a pay-per-use basis from the cloud service providers to deploy a Spark cluster entirely on the cloud. Nevertheless, using only cloud resources can be costly. Hence, both local and cloud resources nowadays are used together to deploy a hybrid cloud computing cluster. However, scheduling jobs in a cluster deployed on hybrid clouds is challenging in the presence of various Service-Level Agreement (SLA) demands such as cost minimization and job deadline guarantee. Most of the existing works either consider a public or a locally deployed cluster and mainly focus on improving job performance in the cluster. In this article, we propose efficient scheduling algorithms that leverage from different VM instance pricing in a hybrid cloud deployed cluster to optimize the Virtual Machine (VM) usage cost for both local and cloud resources and maximize the job deadline met percentage. We have conducted extensive simulation-based experiments to compare our proposed algorithms with the baseline approaches. In addition, we have developed a prototype system on top of Apache Mesos cluster manager and performed real experiments to evaluate the applicability of our proposed approaches in a real platform with benchmark applications. The results show that our proposed algorithms are highly scalable and reduce the cost of VM usage of a hybrid cluster for up to 20 percent. [ABSTRACT FROM AUTHOR]

Recent applications in low-power (1-20 mW) near-sensor computing require the adoption of floating-point arithmetic to reconcile high precision results with a wide dynamic range. In this article, we propose a low-power multi-core computing cluster that leverages the fined-grained tunable principles of transprecision computing to provide support to near-sensor applications at a minimum power budget. Our solution – based on the open-source RISC-V architecture – combines parallelization and sub-word vectorization with a dedicated interconnect design capable of sharing floating-point units (FPUs) among the cores. On top of this architecture, we provide a full-fledged software stack support, including a parallel low-level runtime, a compilation toolchain, and a high-level programming model, with the aim to support the development of end-to-end applications. We performed an exhaustive exploration of the design space of the transprecision cluster on a cycle-accurate FPGA emulator, varying the number of cores and FPUs to maximize performance. Orthogonally, we performed a vertical exploration to identify the most efficient solutions in terms of non-functional requirements (operating frequency, power, and area). We conducted an experimental assessment on a set of benchmarks representative of the near-sensor processing domain, complementing the timing results with a post place-&-route analysis of the power consumption. A comparison with the state-of-the-art shows that our solution outperforms the competitors in energy efficiency, reaching a peak of 97 Gflop/s/W on single-precision scalars and 162 Gflop/s/W on half-precision vectors. Finally, a real-life use case demonstrates the effectiveness of our approach in fulfilling accuracy constraints. [ABSTRACT FROM AUTHOR]

Power and energy is the first-class design constraint for multi-core processors and is a limiting factor for future-generation supercomputers. While modern processor design provides a wide range of mechanisms for power and energy optimization, it remains unclear how software can make the best use of them. This article presents a novel approach for runtime power optimization on modern multi-core systems. Our policy combines power capping and uncore frequency scaling to match the hardware power profile to the dynamically changing program behavior at runtime. We achieve this by employing reinforcement learning (RL) to automatically explore the energy-performance optimization space from training programs, learning the subtle relationships between the hardware power profile, the program characteristics, power consumption and program running times. Our RL framework then uses the learned knowledge to adapt the chip's power budget and uncore frequency to match the changing program phases for any new, previously unseen program. We evaluate our approach on two computing clusters by applying our techniques to 11 parallel programs that were not seen by our RL framework at the training stage. Experimental results show that our approach can reduce the system-level energy consumption by 12 percent, on average, with less than 3 percent of slowdown on the application performance. By lowering the uncore frequency to leave more energy budget to allow the processor cores to run at a higher frequency, our approach can reduce the energy consumption by up to 17 percent while improving the application performance by 5 percent for specific workloads. [ABSTRACT FROM AUTHOR]

Noisy gate-model quantum processing units (QPUs) are currently available from vendors over the cloud, and digital quantum programming approaches exist to run low-depth circuits on physical hardware. These digital representations are ultimately lowered to pulse-level instructions by vendor quantum control systems to affect unitary evolution representative of the submitted digital circuit. Vendors are beginning to open this pulse-level control system to the public via specified interfaces. Robust programming methodologies, software frameworks, and backend simulation technologies for this analog model of quantum computation will prove critical to advancing pulse-level control research and development. Prototypical use cases for this include error mitigation, optimal pulse control, and physics-inspired pulse construction. Here we present an extension to the XACC quantum-classical software framework that enables pulse-level programming for superconducting, gate-model quantum computers, and a novel, general, and extensible pulse-level simulation backend for XACC that scales on classical compute clusters via MPI. Our work enables custom backend Hamiltonian definitions and gate-level compilation to available pulses with a focus on performance and scalability. We end with a demonstration of this capability, and show how to use XACC for pertinent pulse-level programming tasks. [ABSTRACT FROM AUTHOR]

For many data-parallel computing systems like Spark, a job usually consists of multiple computation stages and inter-stage communication (i.e., coflows). Many efforts have been done to schedule coflows and jobs independently. The simple combination of coflow scheduling and job scheduling, however, would prolong the average job completion time (JCT) due to the conflict. For this reason, we propose a new abstraction of scheduling unit, named coBranch, which takes the dependency between computation stages and coflows into consideration, to schedule coflows and jobs jointly. Besides, mainstream coflow schedulers are order-preserving, i.e., all coflows of a high-priority job are prioritized than those of a low-priority job. We observe that the order-preserving constraint incurs low inter-job parallelism. To overcome the problem, we employ an urgency-based mechanism to schedule coBranches, which aims to decrease the average JCT by enhancing the inter-job parallelism. We implement the urgency-based coBranch Scheduling (BS) method on Apache Spark, conduct prototype-based experiments, and evaluate the performance of our method against the shortest-job-first critical-path method and the FIFO method. Results show that our method achieves around 10 and 15 percent reduction in the average JCT, respectively. Large-scale simulations based on the Google trace show that our method performs better and reduces JCT by 23 and 35 percent, respectively. [ABSTRACT FROM AUTHOR]

Distributed synchronous stochastic gradient descent has been widely used to train deep neural networks (DNNs) on computer clusters. With the increase of computational power, network communications generally limit the system scalability. Wait-free backpropagation (WFBP) is a popular solution to overlap communications with computations during the training process. In this article, we observe that many DNNs have a large number of layers with only a small amount of data to be communicated at each layer in distributed training, which could make WFBP inefficient. Based on the fact that merging some short communication tasks into a single one can reduce the overall communication time, we formulate an optimization problem to minimize the training time in pipelining communications and computations. We derive an optimal solution that can be solved efficiently without affecting the training performance. We then apply the solution to propose a distributed training algorithm named merged-gradient WFBP (MG-WFBP) and implement it in two platforms Caffe and PyTorch. Extensive experiments in three GPU clusters are conducted to verify the effectiveness of MG-WFBP. We further exploit trace-based simulations of 4 to 2048 GPUs to explore the potential scaling efficiency of MG-WFBP. Experimental results show that MG-WFBP achieves much better scaling performance than existing methods. [ABSTRACT FROM AUTHOR]