Adaptive Scheduling Framework Research Articles

Adaptive Scheduling Framework of Streaming Applications based on Resource Demand Prediction with Hybrid Algorithms

Adaptive Scheduling Framework of Streaming Applications based on Resource Demand Prediction with Hybrid Algorithms

Adaptive scheduling of mixed bus services with flexible fleet size assignment under demand uncertainty

This paper presents an optimization framework for adaptive scheduling of mixed bus services with flexible fleet size assignment under demand uncertainty. The service scheduling plans are driven by prevailing stochastic passenger demand subject to operational constraints. The optimization problem is formulated as a Markov decision process which aims to minimize passengers’ in-vehicle and waiting times, as well as the operator’s cost via use of services with different routes, schedules, and fleet sizes. To address the computational challenges, the solutions to the problem are to be calculated with use of reinforcement learning techniques. The proposed framework is implemented and tested using real-world scenarios configured from actual bus service route in Hong Kong. Experiment results reveals the benefits of the proposed bus service scheduling framework with use of flexible routes and fleet sizes in saving passengers’ and operator’s costs. This study contributes to real time transit operational planning with advanced computing and optimization techniques.

Theoretical Validation and Hardware Implementation of Dynamic Adaptive Scheduling for Heterogeneous Systems on Chip

Domain-specific systems on chip (DSSoCs) aim to narrow the gap between general-purpose processors and application-specific designs. CPU clusters enable programmability, whereas hardware accelerators tailored to the target domain minimize task execution times and power consumption. Traditional operating system (OS) schedulers can diminish the potential of DSSoCs, as their execution times can be orders of magnitude larger than the task execution time. To address this problem, we propose a dynamic adaptive scheduling (DAS) framework that combines the advantages of a fast, low-overhead scheduler and a sophisticated, high-performance scheduler with a larger overhead. We present a novel runtime classifier that chooses the better scheduler type as a function of the system workload, leading to improved system performance and energy-delay product (EDP). Experiments with five real-world streaming applications indicate that DAS consistently outperforms fast, low-overhead, and slow, sophisticated schedulers. DAS achieves a 1.29× speedup and a 45% lower EDP than the sophisticated scheduler under low data rates and a 1.28× speedup and a 37% lower EDP than the fast scheduler when the workload complexity increases. Furthermore, we demonstrate that the superior performance of the DAS framework also applies to hardware platforms, with up to a 48% and 52% reduction in the execution time and EDP, respectively.

An intelligent scheduling framework for DNN task acceleration in heterogeneous edge networks

With the upgrade of hardware architecture and device capacities, many accelerator-based hardware platforms have been widely deployed in Mobile Edge Computing (MEC) environments. The execution time of many computation-intensive applications (e.g., face recognition and pedestrian detection) can be significantly reduced when deployed on these heterogeneous devices. Moreover, thanks to the popularity of Deep Learning (DL), most terminal applications are integrated with Deep Neural Networks (DNN) and can be divided into interdependent tasks. The structure of these applications can be represented as the Directed Acyclic Graph (DAG). Therefore, it is critical to seek the optimal scheduling order and execution placement of tasks according to the acceleration effects of edge servers and the task dependency. However, conventional scheduling strategies focus on the short-term performance, potentially leading to service quality degradation in the long term. Besides, many studies use Deep Reinforcement Learning (DRL) algorithms to seek a long-term optimal scheduling strategy but ignore the device acceleration and the task dependency. Furthermore, training a well-performed DRL agent is time-consuming, and the large scale of trial-and-error will take up tremendous computation and storage resources. In this paper, we model the scheduling process as a Markov Decision Process (MDP) and design an adaptive scheduling framework for task acceleration. Fully considering the data dependencies, resource conditions, and network conditions, the proposed scheduling algorithm called Meta-AC uses policy gradient combined with meta-learning to minimize the average task delay and the ratio of time-out tasks. As a hierarchical DRL algorithm, Meta-AC uses meta data to learn directed exploration strategies in the high-level agent, improving the learning efficiency from experience samples. Extensive simulations demonstrate the superiority of the proposed method over the counterpart methods.

EALI: Energy-aware layer-level scheduling for convolutional neural network inference services on GPUs

The success of convolutional neural networks (CNNs) has made low-latency inference services on Graphic Processing Units (GPUs) a hot research topic. However, GPUs are hardware processors with high power consumption. To have the least energy consumption while meeting latency Service-Level-Objective (SLO), batching strategy and dynamic voltage frequency scaling (DVFS) are two important solutions. However, existing studies do not coordinate them and regard CNN as a black box, which makes inference services less energy-efficient. In this paper, we propose EALI, an energy-aware layer-level adaptive scheduling framework that is comprised of a power prediction model, a layer combination strategy, and an energy-aware layer-level scheduler. The power prediction model uses classic machine learning techniques to predict fine-grained layer-level power consumption. The layer combination strategy combines multiple layers into optimization units to lower scheduling overhead and complexity. The energy-aware layer-level scheduler adaptively coordinates batching strategy and layer-level DVFS according to workloads to minimize the energy consumption while meeting SLO. Our experimental results on NVIDIA Tesla M40 and V100 GPUs show that, compared to the state-of-the-art approaches, EALI decreases energy consumption by up to 36.24% while meeting SLO.

DAS: Dynamic Adaptive Scheduling for Energy-Efficient Heterogeneous SoCs

Domain-specific systems-on-chip (DSSoCs) aim at bridging the gap between application-specific integrated circuits (ASICs) and general-purpose processors. Traditional operating system (OS) schedulers can undermine the potential of DSSoCs since their execution times can be orders of magnitude larger than the execution time of the task itself. To address this problem, we propose a dynamic adaptive scheduling (DAS) framework that combines the benefits of a fast (low-overhead) scheduler and a slow (sophisticated, high-performance but high-overhead) scheduler. Experiments with five real-world streaming applications show that DAS consistently outperforms both the fast and slow schedulers. For 40 different workloads, DAS achieves on average 1.29x speedup and 45% lower EDP compared to the sophisticated scheduler at low data rates and 1.28x speedup and 37% lower EDP than the fast scheduler when the workload complexity increases.

EAIS: Energy-aware adaptive scheduling for CNN inference on high-performance GPUs

Recently, a large number of convolutional neural network (CNN) inference services have emerged on high-performance Graphic Processing Units (GPUs). However, GPUs are high power consumption units, and the energy consumption increases sharply along with the deployment of deep learning tasks. Although previous studies have considered the latency Service-Level-Objective (SLO) of inference services, they fail to directly take account of the energy consumption. Our investigation shows that coordinating batching and dynamic voltage frequency scaling (DVFS) settings can decrease the energy consumption of CNN inference. But it is affected by (i) larger configuration spaces; (ii) GPUs’ underutilization while data are transferred between CPUs and GPUs; (iii) fluctuating workloads. In this paper, we propose EAIS, an energy-aware adaptive scheduling framework that is comprised of a performance model, an asynchronous execution strategy, and an energy-aware scheduler. The performance model provides valid information about the performance characteristics of CNN inference services to shrink the feasible configuration space. The asynchronous execution strategy overlaps data upload and GPU execution to improve the system processing capacity. The energy-aware scheduler adaptively coordinates batching and DVFS according to fluctuating workloads to minimize energy consumption while meeting latency SLO. Our experimental results on NVIDIA Tesla M40 and V100 GPUs show that, compared to the state-of-the-art methods, EAIS decreases the energy consumption by up to 28.02% and improves the system processing capacity by up to 7.22% while meeting latency SLO. Besides, EAIS has been proved to have good versatility under different latency SLO constraints.

A Framework for Adaptive Scheduling in Cellular Manufacturing Systems

The Fourth Industrial Revolution has made efficient and adaptive manufacturing a prominent research topic. The necessity for personalized products leads to complex production lines. Due to the increased complexity, planning and scheduling of manufacturing processes seeks to identify near-optimum solutions to ensure fast and precise decision making. This research paper aims to contribute to the field of adaptive scheduling by proposing an algorithm that enables near real-time cooperation among machines, workforce and the production manager. A framework for self-adaptive scheduling in a real-world manufacturing case deriving from an SME that produces solar panels will be presented.

Cloud Resource Adaptive Scheduling Framework and Optimization Strategy Based on Swarm Intelligence

Resource scheduling framework is very important for the overall cloud service efficiency and service quality,which is the core module of cloud computing. This paper proposes an adaptive cloud computing resource scheduling framework based on swarm intelligence. The proposed adaptive framework includes resource deployment module, scheduling module, recommendation module, optimization module, and monitoring module.The simulation results show that the proposed framework is more efficient than the traditional framework in terms of task execution time, system load balancing and resource service quality, and can improve the service quality of cloud applications while improving resource utilization.

Scheduling-Efficient Framework for Neural Network on Heterogeneous Distributed Systems and Mobile Edge Computing Systems

As the volume of machine learning training data sets and the quantity of model parameters continue to grow, the pattern in which machine learning models are trained alone can no longer accommodate large-scale data environments. However, distributed systems and mobile edge computing systems are unpredictable and have heterogeneous nodes, resulting in interruptions in training or low convergence rate. In addition, existing distributed machine learning frameworks cannot guarantee a good convergence rate and speedup ratio in a variety of operating environments. Considering the above shortcomings, this paper proposes an adaptive scheduling framework for machine learning based on a heterogeneous distributed system and mobile edge computing system for machine learning model optimization. The framework detects and analyzes the dynamic changes of resources in the distributed system and mobile edge computing system through the resource detection system; then, the task scheduling system adaptively modifies the environmental parameters and schedules calculations. Relevant experiments conducted with the public data set show that the robustness and scalability of the framework are significantly better than the traditional distributed machine learning framework under the premise of ensuring high convergence rate.

A multi-policy adaptive scheduling framework in virtual clouds

With the development of cloud computing technology, many cloudbased data centres are beginning to connect with each other aiming at building large-scale federated cloud platforms. In such a federated cloud platform, the efficiency and effectiveness of traditional schedulers will be significantly degraded due to the unpredictable resource availability and high networking latency. In this paper, we propose a multi-policy adaptive scheduling framework, which is capable of evaluating the scheduling scheme made by existing schedulers. In this way, the proposed scheduling framework can make full use of the advantages of existing scheduling policies and avoid their shortcomings, and therefore makes the final scheduling decisions more efficient and effectiveness. In addition, the proposed scheduling framework is designed as extensible component that can be easily extended by incorporating other scheduling policies. The prototype of this scheduling framework is tested in a real-world federated cloud platform. The experimental results show that it can significantly improve the quality-of-service for the underlying cloud infrastructures as well as cloud user's satisfactory.

A multi-policy adaptive scheduling framework in virtual clouds

A multi-policy adaptive scheduling framework in virtual clouds

Adaptive Scheduling Framework for Multi - Core Systems Based on the Task - Parallel Programming Model

Adaptive Scheduling Framework for Multi - Core Systems Based on the Task - Parallel Programming Model

Adaptive Scheduling Framework for Real-Time Video Encoding on Heterogeneous Systems

To challenge real-time encoding of high-definition video sequences on heterogeneous desktop systems, a collaborative central processing units (CPU) + graphics processing unit (GPU) framework for interloop video encoding is proposed herein. The proposed framework considers the overall complexity of the collaborative interloop encoding as a unified optimization problem. Several functional blocks are integrated for simultaneous execution control, automatic data access management, performance characterization, and adaptive scheduling and load balancing. These blocks aim at fully exploiting the performance of heterogeneous devices, asymmetric bandwidth of communication links, and several levels of concurrency between computation and communication. To support a wide range of CPU and GPU architectures, a specific encoding library is developed with highly optimized algorithms for all interloop modules. The experimental results show that the proposed framework allows achieving a real-time encoding of full high-definition sequences in several CPU + GPU systems. It also delivers performance improvements of up to 61.2% over the state-of-the-art solution, while outperforming individual GPU and quad-core CPU executions by more than 2 and 5 times, respectively.

Near-Optimal One-Sided Scheduling for Coded Segmented Network Coding

As a variation of random linear network coding, segmented network coding (SNC) has attracted great interest in data dissemination over lossy networks due to its low computational cost. In order to guarantee the success of decoding, SNC can adopt a feedbackless forward error correction (FEC) approach by applying a linear block code to the input packets before segmentation at the source node. In particular, if the empirical rank distribution of transfer matrices of segments is known in advance, several classes of coded SNC can achieve close-to-optimal decoding performance. However, the empirical rank distribution in the absence of feedback has been little investigated yet, making the whole performance of the FEC approach unknown. To close this gap, in this paper, we present the first comprehensive study on the transmission scheduling issue for the FEC approach, aiming at optimizing the rank distribution of transfer matrices with little control overhead. We propose an efficient adaptive scheduling framework for coded SNC in lossy unicast networks. This framework is one-sided (i.e., each network node forwards the segments adaptively only according to its own state) and scalable (i.e., its buffer cost will not keep on growing when the number of input packets goes to infinity). The performance of the framework is further optimized based on a linear programming approach. Extensive numerical results show that our framework performs near-optimally with respect to the empirical rank distribution.

SLA Constrained Adaptive Scheduling of Parallel Jobs in a Computational Desktop Grid

Desktop grids are usually equipped with hundreds or thousands of desktops which use the idle cycles of desktop PCs of small enterprises and institutions. From conventional multi -site cluster grids, desktop grids vary markedly in terms of their dynamic nature. In terms of their dynamic nature such grids vary markedly from conventional grids] .This causes the need of new scheduling algorithms ,tailor-made for such systems .Such desktop grids become worthwhile research issue since the nature of these grids attract highly parallel algorithms ,designing efficient scheduling algorithms of parallel jobs on desktop grids .A SLA based adaptive scheduling framework (ASF) for desktop grids has been presented in this paper ,where based on a resource selection algorithm after SLA verification that set by the users tasks are assigned to different available grid resources .Based on a given set of parameter and previous execution log in an adaptive(flexible) manner the resource selection algorithm ,which works in an online mode, selects resources .Whether the user submitted jobs can be executed within the deadline and other constraints specified in the SLA or not the execution log decides .Choosing the most relevant computational parameter depending upon the tasks that are assigned to grid resources is also decided by it during task execution. Later, if performance of any task degrades during the task execution, then adaptive algorithm assists the scheduling policy to adapt the system to achieve the high throughput by re-scheduling to either the server node or to the best available local node at present. Unpredictable execution conditions commonly encountered on desktop grids which can violate the SLA constraints set by the user can also be handled by this method. In this paper, for setting up desktop grid test bed GridGain 2.0 has been used and the performance of the aforesaid SLA based adaptive scheduling framework has been presented.

An Adaptive Scheduling Algorithm Using Period Adjustment

Real-time system technology traditionally developed for safety critical system, has now been extended to support multimedia system and virtual reality. A large number of real-time applications, related to multimedia and adaptive control system, require more flexibility than classical real-time theory usually permits. This paper proposes an efficient adaptive scheduling framework in real-time system based on period adjustment. Under this model periodic tasks can change their execution rates based on their importance values to keep the system underloaded.

Profit-oriented resource allocation using online scheduling in flexible heterogeneous networks

In this paper, we discuss a generalized measurement-based adaptive scheduling framework for dynamic resource allocation in flexible heterogeneous networks, in order to ensure efficient service level performance under inherently variable traffic conditions. We formulate our generalized optimization model based on the notion of a profit center with an arbitrary number of service classes, nonlinear revenue and cost functions and general performance constraints. Subsequently, and under the assumption of a linear pricing model and average queue delay requirements, we develop a fast, low complexity algorithm for online dynamic resource allocation, and examine its properties. Finally, the proposed scheme is validated through an extensive simulation study.

APEX: adaptive disk scheduling framework with QoS support

APEX is an adaptive disk scheduling framework with Quality-of-Service (QoS) support designed for environments with highly varying disk bandwidth usage. APEX is based on a three-layer scheduling architecture: (1) the upper layer realizes different service classes using a set of queues; (2) the mid-layer distributes available disk bandwidth among these queues; and (3) the lower layer is handled by the disk itself, which does the final ordering of disk requests. We demonstrate the use of APEX in an example scenario, a Learning-on-Demand (LoD) application supported by a multimedia system, where students can search for and playback multimedia-based learning material. In this paper, we present the scheduling concepts of APEX which are based on an extended token bucket algorithm. The disk requests scheduled for service are assembled into batches in order to exploit the intelligence of modern disks. Combined with a specialized work-conservation scheme, this enables APEX to apply bandwidth where it is needed, without the loss of efficiency. We demonstrate, through simulations, that APEX provides both higher throughput and lower response times than other mixed-media disk schedulers while still avoiding deadline violations for real-time requests. We also show its robustness with respect to misaligned bandwidth allocation.

An adaptive scheduling framework for heterogeneous computer networks

This paper presents an adaptive scheduling framework for large-scale computer networks facing significant environmental dynamics such as shifts in job mix and available resources. The presented framework detects changes in the job stream, and other events, and evolves scheduling policies at two distinct levels—operational and tactical—in response to the changes, making adaptations such as reassigning resources and adjusting sequencing heuristics. The paper also presents a simulation of a computer network facing dynamic job arrival rates, which demonstrates that even a simple implementation of the framework can lead to major gains in performance.