Heterogeneous Computational Resources Research Articles

Toward a new linpack‐like benchmark for heterogeneous computing resources

SummaryThis work describes some first efforts to design a new Linpack‐like benchmark useful to evaluate the performance of Heterogeneous Computing Resources. The benchmark is based on the Schur Complement reformulation of the solution of a linear equation system. Details about its implementation and evaluation, mainly in terms of performance scalability, are presented for a computing environment based on multi NVIDIA GP‐GPUs nodes connected by an Infiniband network.

Application-Aware Resource Allocation Based on Benefit–Cost Ratio in Computing Power Network with Heterogeneous Computing Resources

The computing power network (CPN) is expected to realize the efficient provisioning of heterogeneous computing power through the collaboration between cloud computing and edge computing. Heterogeneous computing resources consist of CPU, GPU, and other types of computing power. Different types of applications may have diverse requirements for heterogeneous computing resources, such as general applications, CPU-intensive applications, and GPU-intensive applications. Service providers are concerned about how to dynamically provide heterogeneous computing resources for different applications in a cost-effective manner, and how to deploy more applications as much as possible with limited resources. In this paper, the concept of the benefit–cost ratio (BCR) is proposed to quantify the usage efficiency of CPU and GPU in CPNs. An application-aware resource allocation (AARA) algorithm is designed for processing different types of applications. With massive simulations, we compare the performance of the AARA algorithm with a benchmark. In terms of blocking probability, resource utilization, and BCR, AARA achieves better performance than the benchmark. The simulation results indicate that more computing tasks can be accommodated by reducing 3.7% blocking probability through BCR-based resource allocation.

Runtime Resource Management with Multiple-Step-Ahead Workload Prediction

Modern embedded platforms need sophisticated resource managers to utilize their heterogeneous computational resources efficiently. Furthermore, such platforms are subject to fluctuating workloads that are unforeseeable at design time. Predicting the incoming workload could enhance the efficiency of resource management in this situation. But is that the case? And, if so, how substantial is this improvement? Does multiple-step-ahead prediction of the workload contribute to this improvement? How precise must the prediction be to improve decisions rather than cause harm? By proposing a prediction-based resource manager that aims at meeting task deadlines while minimizing energy usage, and by conducting extensive tests, we attempt to provide answers to the aforementioned questions.

The Design and Implementation of a Steganographic Communication System over In-Band Acoustical Channels

This article presents SoundSticker, a system for steganographic, in-band data communication over an acoustic channel. In contrast with recent works that hide bits in inaudible frequency bands, SoundSticker embeds hidden bits in the audible sounds, making them more reliably survive audio codecs and bandpass filtering, while achieving a higher data rate and remaining imperceptible to a listener. The key observation behind SoundSticker is that the human ear is less sensitive to the audio phase changes than the frequency and amplitude changes, which leaves us an opportunity to alter the phase of an audio clip to convey hidden information. We take advantage of this opportunity and build an OFDM-based physical layer. To make this PHY-layer design work for a variety of end devices with heterogeneous computation resources, SoundSticker addresses multiple technical challenges including perceivable waveform artifacts caused by the phase-based modulation, bit rate adaptation without channel sounding and real-time preamble detection. Our prototype on both smartphones and ESP32 platforms demonstrates SoundSticker’s superior performance against the state of the arts, while preserving excellent sound quality and remaining unaffected by common audio codecs like MP3 and AAC. Audio clips produced by SoundSticker can be found at https://soundsticker.github.io/ .

Chronos: Accelerating Federated Learning With Resource Aware Training Volume Tuning at Network Edges

Due to the limited resources and data privacy issue, last decade witnesses the fast development of Distributed Machine Learning (DML) at network edges. Among all the existing DML paradigms, Federated Learning (FL) would be a promising one, since in FL, each client trains its local model without sharing the raw data with others. A community of clients with the same interest can join together to derive a high-performance model by periodically synchronizing the parameters of their local models under the help of a coordination server. However, FL will encounter the straggler problem at network edges, and hence the synchronization among clients becomes inefficient. It slows down the convergence speed of learning process. To alleviate the straggler problem, we propose a method named Chronos that accelerates FL with training volume tuning in this paper. More specifically, Chronos is a resource aware method that adaptively adjusts the amount of data used by each client for training (i.e., training volume) in each iteration in order to eliminate the synchronization waiting time caused by the heterogeneous and dynamical computing and communication resources. In addition, we theoretically analyze the convergence of Chronos in a non-convex setting and utilize the results for the algorithm design of Chronos in return to guarantee the convergence. Extensive experiments show that compared with the benchmark algorithms (i.e., BSP and SSP), Chronos significantly improves convergence speed by up to 6.4×.

A Proposed Scheduling Algorithm for IoT Applications in a Merged Environment of Edge, Fog, and Cloud

With the rapid increase of Internet of Things (IoT) devices and applications, the ordinary cloud computing paradigm soon becomes outdated. Fog computing paradigm extends services provided by a cloud to the edge of network in order to satisfy requirements of IoT applications such as low latency, locality awareness, low network traffic, mobility support, and so forth. Task scheduling in a Cloud-Fog environment plays a great role to assure diverse computational demands are met. However, the quest for an optimal solution for task scheduling in the such environment is exceedingly hard due to diversity of IoT applications, heterogeneity of computational resources, and multiple criteria. This study approaches the task scheduling problem with aims at improving service quality and load balancing in a merged computing system of Edge-Fog-Cloud. We propose a Multi-Objective Scheduling Algorithm (MOSA) that takes into account the job characteristics and utilization of different computational resources. The proposed solution is evaluated in comparison to other existing policies named LB, WRR, and MPSO. Numerical results show that the proposed algorithm improves the average response time while maintaining load balancing in comparison to three existing policies. Obtained results with the use of real workloads validate the outcomes.

Multitask Scheduling Method for Heterogeneous Computing Resources of Power Cryptographic Service

With the rapid development of the power cryptographic service, a single cryptographic computing resource, whether virtual cipher machine (VCM) or physical cipher machine (PCM), is hard to meet the security and efficiency requirements at the same time. Therefore, based on the PCM+VCM hybrid cryptographic computing platform, this brief proposes a novel multitask scheduling algorithm to satisfy the various security requirements and improve the computational efficiency of the power cryptographic service system (PCSS). Firstly, a scheduling architecture and a model for cryptographic service tasks considering security and efficiency are designed. Then, we propose a clustering mapping scheduling (CMS) algorithm based on weighted deep embedding clustering (WDEC) to implement scene partitioning and matching for cryptographic tasks. Moreover, we conduct case studies in a simulation platform that we build with S-9 PCMs and S-1 VCMs. Simulation results reveal that the trade-off between security and efficiency of task sequence is affected by the division of task execution scenarios.

Minimum Overhead Beamforming and Resource Allocation in D2D Edge Networks

Device-to-device (D2D) communications is expected to be a critical enabler of distributed computing in edge networks at scale. A key challenge in providing this capability is the requirement for judicious management of the heterogeneous communication and computation resources that exist at the edge to meet processing needs. In this paper, we develop an optimization methodology that considers the network topology jointly with device and network resource allocation to minimize total D2D overhead, which we quantify in terms of time and energy required for task processing. Variables in our model include task assignment, CPU allocation, subchannel selection, and beamforming design for multiple-input multiple-output (MIMO) wireless devices. We propose two methods to solve the resulting non-convex mixed integer program: semi-exhaustive search optimization, which represents a “best-effort” at obtaining the optimal solution, and efficient alternate optimization, which is more computationally efficient. As a component of these two methods, we develop a novel coordinated beamforming algorithm which we show obtains the optimal beamformer for a common receiver characteristic. Through numerical experiments, we find that our methodology yields substantial improvements in network overhead compared with local computation and partially optimized methods, which validates our joint optimization approach. Further, we find that the efficient alternate optimization scales well with the number of nodes, and thus can be a practical solution for D2D computing in large networks.

Toward a reference architecture based science gateway framework with embedded e‐learning support

AbstractScience gateways have been widely utilized by a large number of user communities to simplify access to complex distributed computing infrastructures. While science gateways are still becoming increasingly popular and the number of user communities is growing, the fast and efficient creation of new science gateways and the flexibility to deploy these gateways on‐demand on heterogeneous computational resources, remain a challenge. Additionally, the increase in the number of users, especially with very different backgrounds, requires intuitive embedded e‐learning tools that support all stakeholders to find related learning material and to guide the learning process. This paper introduces a novel science gateway framework that addresses these challenges. The framework supports the creation, publication, selection, and deployment of cloud‐based reference architectures that can be automatically instantiated and executed even by nontechnical users. The framework also incorporates a knowledge repository exchange and learning module that provides embedded e‐learning support. To demonstrate the feasibility of the proposed solution, two scientific case studies are presented based on the requirements of the plasmasphere, ionosphere, and thermosphere research communities.

Offline Joint Network and Computational Resource Allocation for Energy-Efficient 5G and beyond Networks

In order to cope with the ever-increasing traffic demands and stringent latency constraints, next generation, i.e., sixth generation (6G) networks, are expected to leverage Network Function Virtualization (NFV) as an enabler for enhanced network flexibility. In such a setup, in addition to the traditional problems of user association and traffic routing, Virtual Network Function (VNF) placement needs to be jointly considered. To that end, in this paper, we focus on the joint network and computational resource allocation, targeting low network power consumption while satisfying the Service Function Chain (SFC), throughput, and delay requirements. Unlike the State-of-the-Art (SoA), we also take into account the Access Network (AN), while formulating the problem as a general Mixed Integer Linear Program (MILP). Due to the high complexity of the proposed optimal solution, we also propose a low-complexity energy-efficient resource allocation algorithm, which was shown to significantly outperform the SoA, by achieving up to 78% of the optimal energy efficiency with up to 742 times lower complexity. Finally, we describe an Orchestration Framework for the automated orchestration of vertical-driven services in Network Slices and describe how it encompasses the proposed algorithm towards optimized provisioning of heterogeneous computation and network resources across multiple network segments.

Managing Time-Sensitive IoT Applications via Dynamic Application Task Distribution and Adaptation

The recent proliferation of the Internet of Things has led to the pervasion of networked IoT devices such as sensors, video cameras, mobile phones, and industrial machines. This has fueled the growth of Time-Sensitive IoT (TS-IoT) applications that must complete the tasks of (1) collecting sensor observations they need from appropriate IoT devices and (2) analyzing the data within application-specific time-bounds. If this is not achieved, the value of these applications and the results they produce depreciates. At present, TS-IoT applications are executed in a distributed IoT environment that consists of heterogeneous computing and networking resources. Due to the heterogeneous and volatile nature (e.g., unpredictable data rates and sudden disconnections) of the IoT environment, it has become a major challenge to ensure the time-bounds of TS-IoT applications. Many existing task management techniques (i.e., techniques that are used to manage the execution of IoT applications in distributed computing resources) that have been proposed to support TS-IoT applications to meet their time-bounds do not provide a sophisticated and complete solution to manage the TS-IoT applications in a manner in which their time-bounds are guaranteed. This paper proposes TIDA, a comprehensive platform for managing TS-IoT applications that includes a task management technique, called DTDA, which incorporates novel task sizing, distribution, and dynamic adaptation techniques. DTDA’s task sizing technique measures the computing resources required to complete each task of the TS-IoT application at hand in each available IoT device, edge computer (e.g., network gateways), and cloud virtual machine. DTDA’s task distribution technique distributes and executes the tasks of each TS-IoT application in a manner that their time-bound requirements are met. Finally, DTDA includes a task adaptation technique that dynamically adapts the distribution of tasks (i.e., redistributes TS-IoT application tasks) when it detects a potential application time-bound violation. The paper describes a proof-of-concept implementation of TIDA that uses Microsoft’s Orleans Actor Framework. Finally, the paper demonstrates that the DTDA task management technique of TIDA meets the time-bound requirements of TS-IoT applications by presenting an experimental evaluation involving real time-sensitive IoT applications from the smart city domain.

A novel hybrid heuristic-based list scheduling algorithm in heterogeneous cloud computing environment for makespan optimization

An efficient workflow scheduling can potentially exploit heterogeneity of resources in heterogeneous cloud computing (HCC) platform commensurate with variable requirement of dependent tasks in a given workflow. Minimizing the total scheduling length, makespan, is essential for application performance in heterogeneous computing systems especially in cloud computing environment. The problem of scheduling a set of different dependent tasks onto a set of heterogeneous computational resources is a well-known NP-Hard problem. Therefore, no polynomial scheduling algorithm for computing the optimal solution exists. For approximating a solution to this problem many algorithms have been proposed, but majority of them have low efficiency. In this paper, a novel hybrid heuristic-based list scheduling (HH-LiSch) algorithm is presented for solving the dependent task scheduling in HCC systems in a bounded number of the fully connected virtual machines (VMs). The novelty of the current paper is to present the new task priority strategy, find appropriate VM's slot time, and utilize task duplication technique. Two novel task priority strategies are applied to prioritize tasks in an efficient ordered list. Then, during the scheduling process an insertion-based procedure is called to find an appropriate potential slot time for performing task duplication technique. If it works, the task duplication is added to rudimentary scheduling scheme. In this way, the final scheduling is gradually generated. To validate the work, the experiments are based on six real-world scientific workflows and a random task graph (RTG); then, the performance is evaluated in terms of makespan, Schedule Length Ratio (SLR), speedup and efficiency. The simulation results prove a significant improvement against other counterparts in literature.

Integration of Distributed Heterogeneous Computing Resources for the MPD Experiment with DIRAC Interware

Integration of Distributed Heterogeneous Computing Resources for the MPD Experiment with DIRAC Interware

Stencil Solvers for PDEs on GPUs: An Example From Cosmology

The increasingly diverse ecosystem of high-performance architectures and programming models presents a mounting challenge for programmers seeking to accelerate scientific computing applications. Code generation offers a promising solution, transforming a simple and general representation of computations into lower level, hardware-specialized and -optimized code. We describe the philosophy set forth by the Python package Pystella, a high-performance framework built upon such tools to solve partial differential equations on structured grids. A hierarchy of abstractions provides increasingly expressive interfaces for specifying physical systems and algorithms. We present an example application from cosmology, using finite-difference methods to solve coupled hyperbolic partial differential equations on (multiple) GPUs. Such an approach may enable a broad range of domain scientists to make efficient use of increasingly heterogeneous computational resources while mitigating the drastic effort and expertise nominally required to do so.

Optimal Job Scheduling With Resource Packing for Heterogeneous Servers

Jobs in modern computing clusters have highly diverse processing duration and heterogeneous resource requirements. In this paper, we consider the problem of job scheduling for a computing cluster comprised of multiple servers with heterogeneous computation resources, while taking the different resource demands of the jobs into account. Our focus is to achieve a low overall job response time for the system (which is also referred to as the job flowtime) while providing fairness between small and large jobs. Since the job flowtime minimization problem under multiple (even homogeneous) servers are known to be NP-hard, we propose an approximation algorithm to tackle the original online scheduling problem by adopting the recently-proposed notion of fractional job flowtime as a surrogate objective for minimization. For the general online job arrival case with multi-dimensional resource requirements, we apply Online Convex Optimization (OCO) techniques to design the corresponding scheduling algorithm with performance guarantees. In the single-dimensional resource setting, we show that the dynamic fit of the online version of our approximate algorithm grows only sublinearly with respect to time and derive a bound for its dynamic regret when comparing to its offline counterpart. While the baseline version of our proposed scheduling algorithm assumes the possibilities of job preemption and job migration across different servers, we show that the extent of job preemption and migration can be well controlled by augmenting the objective function with the corresponding switching costs.

Heterogeneous model parallelism for deep neural networks

Deep neural networks (DNNs) have transformed computer vision, establishing themselves as the current state-of-the-art for image processing. Nevertheless, the training of current large DNN models is one of the main challenges to be solved. In this sense, data-parallelism has been the most widespread distributed training strategy since it is easy to program and can be applied to almost all cases. However, this solution suffers from several limitations, such as its high communication requirements and the memory constraints when training very large models. To overcome these limitations model-parallelism has been proposed, solving the most substantial problems of the former strategy. However, describing and implementing the parallelization of the training of a DNN model across a set of processes deployed on several devices is a challenging task. Current proposed solutions assume a homogeneous distribution, being impractical when working with devices of different computational capabilities, which is quite common on high performance computing platforms. To address previous shortcomings, this work proposes a novel model-parallelism technique considering heterogeneous platforms, where a load balancing mechanism between uneven devices of an HPC platform has been implemented. Our proposal takes advantage of the Google Brain’s Mesh-TensorFlow for convolutional networks, splitting computing tensors across filter dimension in order to balance the computational load of the available devices. Conducted experiments show an improvement in the exploitation of heterogeneous computational resources, enhancing the training performance. The code is available on: https://github.com/mhaut/HeterogeneusModelDNN.

Computation Offloading in LEO Satellite Networks With Hybrid Cloud and Edge Computing

Low earth orbit (LEO) satellite networks can break through geographical restrictions and achieve global wireless coverage, which is an indispensable choice for future mobile communication systems. In this article, we present a hybrid cloud and edge computing LEO satellite (CECLS) network with a three-tier computation architecture, which can provide ground users with heterogeneous computation resources and enable ground users to obtain computation services around the world. With the CECLS architecture, we investigate the computation offloading decisions to minimize the sum energy consumption of ground users, while satisfying the constraints in terms of the coverage time and the computation capability of each LEO satellite. The considered problem leads to a discrete and nonconvex since the objective function and constraints contain binary variables, which makes it difficult to solve. To address this challenging problem, we convert the original nonconvex problem into a linear programming problem by using the binary variables relaxation method. Then, we propose a distributed algorithm by leveraging the alternating direction method of multipliers (ADMMs) to approximate the optimal solution with low computational complexity. Simulation results show that the proposed algorithm can effectively reduce the total energy consumption of ground users.

SPHERE: A Multi-SoC Architecture for Next-Generation Cyber-Physical Systems Based on Heterogeneous Platforms

This paper presents SPHERE, a project aimed at the realization of an integrated framework to abstract the hardware complexity of interconnected, modern system-on-chips (SoC) and simplify the management of their heterogeneous computational resources. The SPHERE framework leverages hypervisor technology to virtualize computational resources and isolate the behavior of different subsystems running on the same platform, while providing safety, security, and real-time communication mechanisms. The main challenges addressed by SPHERE are discussed in the paper along with a set of new technologies developed in the context of the project. They include isolation mechanisms for mixed-criticality applications, predictable I/O virtualization, the management of time-sensitive networks with heterogeneous traffic flows, and the management of field-programmable gate arrays (FPGA) to provide efficient implementations for cryptography modules, as well as hardware acceleration for deep neural networks. The SPHERE architecture is validated through an autonomous driving use-case.

A Co-Scheduling Framework for DNN Models on Mobile and Edge Devices with Heterogeneous Hardware

With the emergence of more and more powerful chipsets and hardware and the rise of Artificial Intelligence of Things (AIoT), there is a growing trend for bringing Deep Neural Network (DNN) models to empower mobile and edge devices with intelligence such that they can support attractive AI applications in a real-time manner. To leverage heterogeneous computational resources (such as CPU, GPU, DSP, etc.) to effectively and efficiently support the concurrent inference of multiple DNN models on a mobile or edge device, we propose a novel online Co-Scheduling framework based on deep REinforcement Learning, called COSREL. COSREL has the following desirable features: 1) it achieves significant speedup over commonly-used methods by efficiently utilizing all the computational resources on heterogeneous hardware; 2) it leverages emerging Deep Reinforcement Learning (DRL) to make dynamic and wise online scheduling decisions based on system runtime state; 3) it is capable of making a good tradeoff among inference latency, throughput, and energy efficiency; and 4) it makes no changes to given DNN models, thus preserves their accuracies. To evaluate COSREL, we conduct extensive experiments on an off-the-shelf Android smartphone. The experimental results show that COSREL consistently outperforms other baselines in terms of throughput, latency, and energy efficiency.

Federated Learning Over Wireless Networks: Convergence Analysis and Resource Allocation

There is an increasing interest in a fast-growing machine learning technique called Federated Learning, in which the model training is distributed over mobile user equipments (UEs), exploiting UEs' local computation and training data. Despite its advantages in data privacy-preserving, Federated Learning (FL) still has challenges in heterogeneity across UEs' data and physical resources. We first propose a FL algorithm which can handle the heterogeneous UEs' data challenge without further assumptions except strongly convex and smooth loss functions. We provide the convergence rate characterizing the trade-off between local computation rounds of UE to update its local model and global communication rounds to update the FL global model. We then employ the proposed FL algorithm in wireless networks as a resource allocation optimization problem that captures the trade-off between the FL convergence wall clock time and energy consumption of UEs with heterogeneous computing and power resources. Even though the wireless resource allocation problem of FL is non-convex, we exploit this problem's structure to decompose it into three sub-problems and analyze their closed-form solutions as well as insights to problem design. Finally, we illustrate the theoretical analysis for the new algorithm with Tensorflow experiments and extensive numerical results for the wireless resource allocation sub-problems. The experiment results not only verify the theoretical convergence but also show that our proposed algorithm outperforms the vanilla FedAvg algorithm in terms of convergence rate and testing accuracy.