Run-Time Management of a MPSoC Containing FPGA Fabric Tiles

Multimedia applications, like, e.g., 3-D games and video decoders, are typically composed of communicating tasks. Their target embedded computing platforms (e.g., TI OMAP3, IBM Cell) contain multiple heterogeneous processing elements. At application design-time, it is often unknown which applications will execute simultaneously. Hence, resource assignment decisions need to be made by a run-time manager. Run-time assignment of these communicating tasks onto the communication and computation resources of such a multiprocessor platform is a challenging task. In the presence of fine-grain reconfigurable hardware processing elements, the run-time manager also needs to consider the creation of a so-called configuration hierarchy. Instead of executing a dedicated hardware task, the fine-grain reconfigurable hardware fabric hosts a programmable softcore block that, in turn, executes the task functionality. Hence, the next challenge for run-time management is to efficiently handle a configuration hierarchy. This paper details a run-time task assignment heuristic that performs fast and efficient task assignment in a multiprocessor system-on-chip containing fine-grain reconfigurable hardware tiles. In addition, this algorithm is capable of managing a configuration hierarchy. We show that being capable of handling a configuration hierarchy significantly improves the task assignment performance (i.e., success rate and assignment quality). In several cases, adding a configuration hierarchy improves the assignment success rate of the assignment heuristic by 20%.     

Combining Coarse-Grained Software Pipelining with DVS for Scheduling Real-Time Periodic Dependent Tasks on Multi-Core Embedded Systems

In this paper, we combine coarse-grained software pipelining with DVS (Dynamic Voltage/Frequency Scaling) for optimizing energy consumption of stream-based multimedia applications on multi-core embedded systems. By exploiting the potential of multi-core architecture and the characteristic of streaming applications, we propose a two-phase approach to solve the energy minimization problem for periodic dependent tasks on multi-core processors with discrete voltage levels. With our approach, in the first phase, we propose a coarse-grained task-level software pipelining algorithm called RDAG to transform the periodic dependent tasks into a set of independent tasks based on the retiming technique (Leiserson and Saxe, Algorithmica 6:5–35, 1991). In the second phase, we propose two DVS scheduling algorithms for energy minimization. For single-core processors, we propose a pseudo-polynomial algorithm based on dynamic programming that can achieve optimal solution. For multi-core processors, we propose a novel scheduling algorithm called SpringS which works like a spring and can effectively reduce energy consumption by iteratively adjusting task scheduling and voltage selection. We conduct experiments with a set of benchmarks from E3S (Dick 2008) and TGFF () based on the power model of the AMD Mobile Athlon4 DVS processor. The experimental results show that our technique can achieve 12.7% energy saving compared with the algorithms in Zhang et al. (2002) on average.

Mapping Application Requirements to Virtualization-Enabled Software Defined WSN

With the emergence of resource powerful sensor nodes, the concept of WSN virtualization is gaining increasing attention from the research community and the industry. One approach to achieve WSN virtualization is to exploit the capabilities of individual sensor nodes to execute tasks of multiple applications concurrently. In this paper, we consider the problem of task allocation in software-defined WSNs (SD-WSNs), which are distinguished by centralized control plane and programmable data plane. We extend our previous work on this topic, where we proposed the control algorithm which determines suitability of a sensor node for task allocation based on the active routing paths and residual energy in the network. Availability of such information can be easily justified in SD-WSNs. Through extensive simulations, the performance of this strategy has been evaluated and compared with two conventional task allocation approaches, which assume traditional minimum-hop routing. In addition, we analysed performance of more simple software defined networking-based approach, which performs resource allocation by considering only residual energy in the network. The obtained results demonstrate benefits of SD-WSN architecture when it comes to virtualization efficiency, and clarify improvements achieved by mutual correlation of routing and task allocation decisions.     

Scate: A Scalable Time and Energy Aware Actor Task Allocation Algorithm in Wireless Sensor and Actor Networks

In many applications of wireless sensor actor networks (WSANs) that often run in harsh environments, the reduction of completion times of tasks is highly desired. We present a new time‐aware, energy‐aware, and starvation‐free algorithm called Scate for assigning tasks to actors while satisfying the scalability and distribution requirements of WSANs with semi‐automated architecture. The proposed algorithm allows concurrent executions of any mix of small and large tasks and yet prevents probable starvation of tasks. To achieve this, it estimates the completion times of tasks on each available actor and then takes the remaining energies and the current workloads of these actors into account during task assignment to actors. The results of our experiments with a prototyped implementation of Scate show longer network lifetime, shorter makespan of resulting schedules, and more balanced loads on actors compared to when one of the three well‐known task‐scheduling algorithms, namely, the max‐min, min‐min, and opportunistic load balancing algorithms, is used.     

Chameleon: Application-Level Power Management

In this paper, we present Chameleon an application-level power management approach for reducing energy consumption in mobile processors. By using application domain knowledge, as opposed to OS-level or hardware-level inferred knowledge, Chameleon can substantially reduce CPU energy consumption. By exporting the energy management to user-space, designers can design more flexible and easily portable algorithms and systems, and use multiple energy management policies simultaneously. Specifically, we propose a minimal operating system interface that applications use to obtain global knowledge from the kernel in order to make local decisions. We consider three classes of applications soft real-time, interactive and batch and design user level power management strategies for representative applications such as a movie player, a word processor, a web browser, and a batch compiler. Our experiments show that, compared to the traditional system-wide CPU voltage scaling approaches, Chameleon can achieve up to 32-50% energy savings while delivering comparable or better performance to applications. Similarly, Chameleon extracts 9-41% more energy when compared to Grace OS, which uses some application knowledge but operates within the kernel. Further, Chameleon imposes minimal overhead and is effective at scheduling concurrent applications with diverse energy needs.     

Application-directed voltage scaling

Clock (and voltage) scheduling is an important technique to reduce the energy consumption of processors that support voltage scaling. It is difficult, however, to achieve good results using only statistics from the operating system level when applications show bursty (unpredictable) behavior. We take the approach that such applications must be made power-aware and specify their average execution time (AET) and the deadline to the scheduler controlling the clock speed and processor voltage. This paper describes our energy priority scheduling (EPS) algorithm supporting power-aware applications. EPS orders tasks according to how tight their deadlines are and how often tasks overlap. Low-priority tasks are scheduled first, since they can be easily preempted to accommodate for high-priority tasks later. The EPS algorithm does not always yield the optimal schedule, but has a low complexity. We have implemented EPS on a StrongARM-based variable-voltage platform. We conducted experiments with a modified video decoder that estimates the AET of each frame. Measurements show that application-directed voltage scaling reduces processor power consumption with 50% for the bursty video decoder without missing any frame deadlines.     

A novel data size-aware offloading technique for resource provisioning in mobile cloud computing

Mobile device users are involved in social networking, gaming, learning, and even some office work, so the end users expect mobile devices with high-response computing capacities, storage, and high battery power consumption. The data-intensive applications, such as text search, online gaming, and face recognition usage, have tremendously increased. With such high complex applications, there are many issues in mobile devices, namely, fast battery draining, limited power, low storage capacity, and increased energy consumption. The novelty of this work is to strike a balance between time and energy consumption of mobile devices while using data-intensive applications by finding the optimal offloading decisions. This paper proposes a novel efficient Data Size-Aware Offloading Model (DSAOM) for data-intensive applications and to predict the appropriate resource provider for dynamic resource allocation in mobile cloud computing. Based on the data size, the tasks are separated and gradually allocated to the appropriate resource providers for execution. The task is placed into the appropriate resource provider by considering the availability services in the fog nodes or the cloud. The tasks are split into smaller portions for execution in the neighbor fog nodes. To execute the task in the remote side, the offloading decision is made by using the min-cut algorithm by considering the monetary cost of the mobile device. This proposed system achieves low-latency time 13.2% and low response time 14.1% and minimizes 24% of the energy consumption over the existing model. Finally, according to experimental findings, this framework efficiently lowers energy use and improves performance for data-intensive demanding application activities, and the task offloading strategy is effective for intensive offloading requests.     

Task allocation and scheduling in wireless distributed computing networks

Wireless distributed computing (WDC) is an enabling technology that allows radio nodes to cooperate in processing complex computational tasks of an application in a distributed manner. WDC research is being driven by the fact that mobile portable computing devices have limitations in executing complex mobile applications, mainly attributed to their limited resource and functionality. This article focuses on resource allocation in WDC networks, specifically on scheduling and task allocation. In WDC, it is important to schedule communications between the nodes in addition to the allocation of computational tasks to nodes. Communication scheduling and heterogeneity in the operating environment make the WDC resource allocation problem challenging to address. This article presents a task allocation and scheduling algorithm that optimizes both energy consumption and makespan in a heuristic manner. The proposed algorithm uses a comprehensive model of the energy consumption for the execution of tasks and communication between tasks assigned to different radio nodes. The algorithm is tested for three objectives, namely, minimization of makespan, minimization of energy consumption, and minimization of both makespan and energy consumption.     

OKRA: optimal task and resource allocation for energy minimization in mobile edge computing systems

To cope with the computational and energy constraints of mobile devices, Mobile Edge Computing (MEC) has recently emerged as a new paradigm that provides IT and cloud-computing services at mobile network edge in close proximity to mobile devices. This paper investigates the energy consumption problem for mobile devices in a multi-user MEC system with different types of computation tasks, random task arrivals, and unpredictable channel conditions. By jointly considering computation task scheduling, CPU frequency scaling, transmit power allocation and subcarrier bandwidth assignment, we formulate it as a stochastic optimization problem aiming at minimizing the power consumption of mobile devices and to maintain the long-term stability of task queues. By leveraging the Lyapunov optimization technique, we propose an online control algorithm (OKRA) to solve the formulation. We prove that this algorithm is able to provide deterministic worst-case latency guarantee for latency-sensitive computation tasks, and balance a desirable tradeoff between power consumption and system stability by appropriately tuning the control parameter. Extensive simulations are carried out to verify the theoretical analysis, and illustrate the impacts of critical parameters to algorithm performance.

     

Cluster-PSO Based Resource Orchestration for Multi-task Applications in Vehicular Cloud

Vehicular cloud computing (VCC) provides a vehicular user attaching several resources with different types at the same time. Additionally, the vehicular applications especially for big data processing are always complicated and may be decomposed into several fine-grained tasks. When offloading the complicated multi-task application to the vehicular clouds, the task executes individually in terms of its own computation, storage and bandwidth requirement. Different from the task offloading in mobile cloud computing which aims to optimize the energy consumption, the important metric for vehicular users is the application delay. Moreover, the moving vehicles always have the similar resource properties and may form the solution clusters when finding the resource orchestration policy, which brings an opportunity of improving resource orchestration performance. In this paper, we formulate the VCC resource orchestration as an optimization problem, and propose a cluster-particle swarm optimization (PSO) algorithm to obtain the resource orchestration policy. A fast cluster algorithm is used to divide the solution space and generate sub-swarms for better exploring the orchestration solutions. The experiment results show that the cluster-PSO algorithm can achieve a higher resource orchestration accuracy in an acceptable time comparing to the other PSO algorithms. Especially, when there are more tasks in an application and the vehicle has more optional VCC resources, the performance of the cluster-PSO based resource orchestration is outstanding.     

Cross-Layer Collaborative In-Network Processing in Multihop Wireless Sensor Networks

Emerging wireless sensor network (WSN) applications demand considerable computation capacity for in-network processing. To achieve the required processing capacity, cross-layer collaborative in-network processing among sensors emerges as a promising solution: sensors do not only process information at the application layer, but also synchronize their communication activities to exchange partially processed data for parallel processing. However, scheduling computation and communication events is a challenging problem in WSNs due to limited resource availability and shared communication medium. In this work, an application-independent task mapping and scheduling solution in multihop homogeneous WSNs, multihop task mapping and scheduling (MTMS), is presented that provides real-time guarantees. Using our proposed application model, the multihop channel model, and the communication scheduling algorithm, computation tasks and associated communication events are scheduled simultaneously. The dynamic voltage scaling (DVS) algorithm is presented to further optimize energy consumption. Simulation results show significant performance improvements compared with existing mechanisms in terms of minimizing energy consumption subject to delay constraints     

Energy characterization of mobile devices and applications using power-thermal benchmarks

Power consumption and heat dissipation are the major factors that limit the performance and mobility of battery-powered devices. As they become key elements in the design of mobile devices and their applications, different power and thermal management strategies have been proposed and implemented during the previous years in order to overcome the mobility limitation due to the battery lifetime. A new energy management approach is to build energy-aware applications so that we have knowledge on the consumed energy while the device is running. In this paper we define two new types of benchmarks, called power and thermal benchmark, which are software applications intended for the run-time system level to provide power and thermal characterization. These benchmarks are an easy way for the applications to adapt their execution pattern, in order to finish their tasks both in time and in the battery lifetime.     

Energy minimization for real-time systems with (m,k)-guarantee

Energy consumption and quality of service (QoS) are two primary concerns in the development of today's pervasive computing systems. While most of the current research in energy-aware real-time scheduling has been focused on hard real-time systems, a large number of practical applications and systems exhibit more soft real-time nature. In this paper, we study the problem of minimizing energy for soft real-time systems while providing a QoS guarantee. The QoS requirements are deterministically quantified with the (m,k)-constraints, which require that at least m out of any k consecutive jobs of a task meet their deadlines. In this paper, we propose a hybrid approach to achieve the dual goals of QoS guarantee and energy minimization. We first present the necessary and sufficient schedulability conditions for the static mandatory/optional workload partitioning. Then, we propose to dynamically vary the statically defined mandatory/optional partitions to accommodate dynamic run-time variations while minimizing the energy consumption. The experimental results demonstrate that our proposed techniques outperform previous work significantly in terms of both the energy savings and achieved QoS.     

A Dynamic Voltage Scaling Algorithm for Dynamic Workloads

Dynamic Voltage Scaling (DVS) is a promising method to achieve energy saving by slowing down the processor into multiple frequency levels in battery-operated embedded systems. However, the worst case execution time (WCET) of the tasks scheduled by DVS must be known ahead of time to ensure their schedulability. In reality, a system’s workloads may change significantly without satisfying any prediction. In other words, a task’s WCET may not provide useful information about its future real execution time (RET). This paper presents a novel Dynamic-Mode EDF scheduling algorithm when workloads change significantly. One of the Single-Mode, Dual-Mode, and Three-Mode frequency setting formats can be applied, based on the RET and the accumulated slack at run-time. Only one combination of the number of modes/speeds, speed-switching transition points, and the frequency scaling factor for each mode can lead to the best energy saving. Experimental results show that, given an RET pattern, our Dynamic-Mode DVS algorithm achieves an average 15% energy savings over the traditional two-mode DVS scheme on hard real-time systems. Additionally, we also consider speed-switching or energy transition overhead, and implement a preliminary test of our proposed algorithm. With a less aggressive voltage scaling strategy (fewer speed changes for each job), deadlines can still be strictly satisfied and an average of 14% energy consumption saving over a non-DVS scheme is observed.

Multi-objective optimization for task offloading based on network calculus in fog environments

With the widespread application of wireless communication technology and continuous improvements to Internet of Things (IoT) technology, fog computing architecture composed of edge, fog, and cloud layers have become a research hotspot. This architecture uses Fog Nodes (FNs) close to users to implement certain cloud functions while compensating for cloud disadvantages. However, because of the limited computing and storage capabilities of a single FN, it is necessary to offload tasks to multiple cooperating FNs for task completion. To effectively and quickly realize task offloading, we use network calculus theory to establish an overall performance model for task offloading in a fog computing environment and propose a Globally Optimal Multi-objective Optimization algorithm for Task Offloading (GOMOTO) based on the performance model. The results show that the proposed model and algorithm can effectively reduce the total delay and total energy consumption of the system and improve the network Quality of Service (QoS).     

基于混合遗传算法的云计算任务节能调度算法

云计算中主机和任务的数量都是十分庞大的,如何通过任务分配调度来减少成本开销和降低能耗是当前云计算和绿色计算领域研究的热点问题。根据云计算任务以及运行环境的特点,将云计算任务分配问题抽象为多维多背包求解问题,并采用改进的混合遗传算法对该问题进行求解。实验结果表明,改进的混合遗传算法能够在较短的时间内找到问题的优化解,并且根据该算法实现的任务分配策略能够有效地减少任务执行的成本开销和能耗。     

Deep reinforcement learning-based joint task offloading and bandwidth allocation for multi-user mobile edge computing

The rapid growth of mobile internet services has yielded a variety of computation-intensive applications such as virtual/augmented reality. Mobile Edge Computing (MEC), which enables mobile terminals to offload computation tasks to servers located at the edge of the cellular networks, has been considered as an efficient approach to relieve the heavy computational burdens and realize an efficient computation offloading. Driven by the consequent requirement for proper resource allocations for computation offloading via MEC, in this paper, we propose a Deep-Q Network (DQN) based task offloading and resource allocation algorithm for the MEC. Specifically, we consider a MEC system in which every mobile terminal has multiple tasks offloaded to the edge server and design a joint task offloading decision and bandwidth allocation optimization to minimize the overall offloading cost in terms of energy cost, computation cost, and delay cost. Although the proposed optimization problem is a mixed integer nonlinear programming in nature, we exploit an emerging DQN technique to solve it. Extensive numerical results show that our proposed DQN-based approach can achieve the near-optimal performance.     

Failure-Aware Task Scheduling of Synchronous Data Flow Graphs Under Real-Time Constraints

As more processors are integrated into Multiprocessor System-on-Chips (MPSoCs) via relentless technology scaling, the mean-time-to-failure (MTTF) is reduced to the extent that unexpected processor failures are considered during design time. A popular approach to tolerate processor failures is to migrate tasks on the faulty processor to live processors. This approach, however, is not suitable for real-time digital signal processing (DSP) applications since it may not guarantee real-time constraints. In this paper, we propose the re-scheduling of the entire application to minimize throughput degradation under a latency constraint, given that the application is specified by a Synchronous Data Flow (SDF) graph. We obtain sub-optimal re-scheduling results using a genetic algorithm for each scenario of processor failures at compile-time. If a failure is detected at run-time, the live processors obtain the saved schedule, perform task transfer, and execute the remaining tasks of the current iteration. We compare preemptive and non-preemptive migration policies and propose a hybrid policy to obtain better performance. We demonstrate the viability of the proposed technique through experiments with real-life DSP applications as well as randomly generated graphs under timing constraints and random fault scenarios.     

Elastic DVS Management in Processors With Discrete Voltage/Frequency Modes

Applying classical dynamic voltage scaling (DVS) techniques to real-time systems running on processors with discrete voltage/frequency modes causes a waste of computational resources. In fact, whenever the ideal speed level computed by the DVS algorithm is not available in the system, to guarantee the feasibility of the task set, the processor speed must be set to the nearest level greater than the optimal one, thus underutilizing the system. Whenever the task set allows a certain degree of flexibility in specifying timing constraints, rate adaptation techniques can be adopted to balance performance (which is a function of task rates) versus energy consumption (which is a function of the processor speed). In this paper, we propose a new method that combines discrete DVS management with elastic scheduling to fully exploit the available computational resources. Depending on the application requirements, the algorithm can be set to improve performance or reduce energy consumption, so enhancing the flexibility of the system. A reclaiming mechanism is also used to take advantage of early completions. To make the proposed approach usable in real-world applications, the task model is enhanced to consider some of the real CPU characteristics, such as discrete voltage/frequency levels, switching overhead, task execution times nonlinear with the frequency, and tasks with different power consumption. Implementation issues and experimental results for the proposed algorithm are also discussed     

Mapping the MPEG-4 visual texture decoder: a system-level design technique based on heterogeneous platforms

This article presents automated techniques supporting the design-time scheduling phase of a unique approach for managing concurrent tasks of dynamic real-time applications mapped on a heterogeneous platform with different types of software and hardware components. This approach is based on design-time exploration, which results in a set of schedules and assignments for each task, represented by Pareto curves. At run-time, a low complexity scheduler selects an optimal combination of working points, exploiting the dynamic and nondeterministic behavior of the system. The combined approach leads to significant overall power savings compared to state-of-the-art dynamic voltage scaling techniques. The design-time generated Pareto curves can also be used by the application designer to effectively make quantitative tradeoffs between system cost and performance.