Application-aware cloudlet selection for computation offloading in multi-cloudlet environment

Latency- and power-aware offloading is a promising issue in the field of mobile cloud computing today. To provide latency-aware offloading, the concept of cloudlet has evolved. However, offloading an application to the most appropriate cloudlet is still a major challenge. This paper has proposed an application-aware cloudlet selection strategy for multi-cloudlet scenario. Different cloudlets are able to process different types of applications. When a request comes from a mobile device for offloading a task, the application type is verified first. According to the application type, the most suitable cloudlet is selected among multiple cloudlets present near the mobile device. By offloading computation using the proposed strategy, the energy consumption of mobile terminals can be reduced as well as latency in application execution can be decreased. Moreover, the proposed strategy can balance the load of the system by distributing the processes to be offloaded in various cloudlets. Consequently, the probability of putting all loads on a single cloudlet can be dealt for load balancing. The proposed algorithm is implemented in the mobile cloud computing laboratory of our university. In the experimental analyses, the sorting and searching processes, numerical operations, game and web service are considered as the tasks to be offloaded to the cloudlets based on the application type. The delays involved in offloading various applications to the cloudlets located at the university laboratory, using proposed algorithm are presented. The mathematical models of total power consumption and delay for the proposed strategy are also developed in this paper.     

云计算环境下基于蜜蜂觅食行为的任务负载均衡算法

针对云计算环境下的任务调度程序通常需要较多响应时间和通信成本的问题,提出了一种基于蜜蜂行为的负载均衡(HBB-LB)算法。首先,利用虚拟机(VM)进行负载平衡来最大化吞吐量;然后,对机器上任务的优先级进行平衡;最后,将平衡重点放在减少VM等待序列中任务的等待时间上,从而提高处理过程的整体吞吐量和优先级。利用CloudSim工具模拟云计算环境进行仿真实验,结果表明,相比粒子群优化(PSO)、蚁群算法(ACO)、动态负载均衡(DLB)、先入先出(FIFO)和加权轮询(WRR)算法, HBB-LB算法的平均响应时间分别节省了5%、13%、17%、67%、37%,最大完成时间分别节省了20%、23%、18%、55%、46%,可以更好地平衡非抢占式独立任务,适用于异构云计算系统。     

基于云模型的被动队列管理算法

为了解决因拥塞而带来的网络性能问题,基于“弃头”方式提出了一种新的被动队列管理算法DFCM。该算法首先结合业务流队长和到达速率给出了丢包策略和丢包概率,同时利用云模型对实际队长进行求解;最后,通过利用NS2和Matlab进行仿真实验,深入分析了影响算法性能的关键因素。与“弃尾”(Drop Tail)和DFSQ算法相比,DFCM在瞬时队长方差、有效传输数据包等方面显示了较好的适应性。     

IEEE 802.11e动态队列分派算法

基于跨层优化和不平等保护的思想,提出一种新的IEEE 802.11e动态队列分派算法。利用AC［1］和AC［0］队列优化视频传输;根据队列状态动态计算队列优先级,并将它与视频帧优先级相匹配;综合考虑视频帧的优先级、不同队列的拥塞状况为视频帧动态分派队列。仿真实验表明,该算法能有效降低视频的传输失真。     

云计算中基于共享机制和群体智能优化算法的任务调度方案

为了提高云计算中虚拟机(VM)的利用率并降低任务的完成时间,提出了一种融合共享机制的混合群智能优化算法,实现云任务的动态调度。首先,将虚拟机调度编码为蜜蜂、蚂蚁和遗传个体。然后,利用人工蜂群算法(ABC)、蚁群算法(ACO)和遗传算法(GA)分别在各自邻域内寻找最优解。最后,通过一个共享机制使3种算法定期交流各自搜索到的解,并将获得的最佳解作为当前最优解进行下一次迭代过程,以此来加速算法收敛并提高收敛精度。通过CloudSim进行了一个云任务调度的仿真实验,结果表明提出的混合算法能够合理有效地调度任务,在任务完成时间和稳定性方面具有优越的性能。     

LATOC: an enhanced load balancing algorithm based on hybrid AHP-TOPSIS and OPSO algorithms in cloud computing

Providing required level of service quality in cloud computing is one of the most significant cloud computing challenges because of software and hardware complexities, different features of tasks and computing resources and also, lack of appropriate distribution of tasks in cloud computing environments. The recent research in this field show that lack of smart prioritization and ordering of tasks in scheduling (as an NP-hard problem) has been very effective and resulted in lack of load balancing, response time increase, total execution time increase and also, average resource use decrease. In line with this, the proposed method of this research called LATOC considered first the key criteria of an input task like required processing unit, data length of task and execution time. Then, it addressed task prioritization in separate queues using the technique for order preference by similarity to ideal solution (TOPSIS) and analytic hierarchy process (AHP) in figure of a hybrid intelligent algorithm (AHP-TOPSIS). Each ordered task in separate priority queues was placed based on its priority level, and then, to assign each task from each priority queue to virtual machines, optimized particle swarm optimization was used. Many simulations based on various scenarios in Cloudsim simulator show that smart assignment of prioritized tasks by LATOC resulted in improvement of important cloud computing parameters such as total execution time and average resource use comparing similar methods.

     

云环境下周期和非周期混合实时任务双容错调度算法

云环境中的处理机故障已成为云计算不可忽视的问题,容错成为设计和发展云计算系统的关键需求。针对一些容错调度算法在任务调度过程中调度效率低下以及任务类型单一的问题,提出一种处理机和任务主副版本分组的容错调度方法;并给出了副版本可重叠执行的判定方法,以及任务最坏响应时间的计算公式。通过实验和分析表明,和以前算法相比,将处理机分成两组分别执行任务主版本和任务副版本,减少了任务调度所需进行可调度测试的时间,增加了副版本重叠执行的机会,减少了所需的处理机个数,对提高系统处理机的利用率和容错调度的效率具有重要的意义。     

Novel efficient on-chip task scheduler for multi-core hard real-time systems

This article presents an efficient hardware architecture of EDF-based task scheduler, which is suitable for hard real-time systems due to the constant response time of the scheduler. The proposed scheduler contains a queue of ready tasks that is based on a new MIN/MAX queue architecture called Heap Queue, which is inspired by Shift Registers, Systolic Arrays, heapsort algorithm, the Rocket Queue architecture and dual-port RAMs. The instructions of the proposed scheduler have throughput of one instruction per two clock cycles regardless of the actual number of tasks managed by the scheduler, and regardless of the scheduler capacity. The developed task scheduler is optimized for low chip area costs, which leads to lower energy consumption. The Heap Queue-based architecture has constant time complexity due to two clock-cycle response time of the instructions and therefore, the architecture is highly deterministic. The scheduler supports CPUs that can execute 1, 2 or 4 tasks simultaneously, and contains an implementation of clever and efficient logic that can handle conflicts caused by the fact that the scheduler is used by all CPU cores at the same time. The proposed scheduler was verified through SystemVerilog UVM-like simulations that applied billions of randomly generated test instructions. Achieved ASIC (28 nm) and FPGA synthesis results are presented and compared. More than 86% of the chip area and 93% of the total power consumption can be saved if Heap Queue architecture is used in hardware implementations of EDF algorithm. Advantages and disadvantages of the proposed task scheduler are discussed through the comparison to the existing solutions.     

大型工程项目任务多目标优化调度方法

提出了一种大型工程项目任务多目标优化调度方法。构建了一种以项目工期最小化、费用最小化及质量最大化为目标函数的多目标优化模型;针对模型的多变量、多约束、大组合量特点,提出了一种基于自适应变异和模拟退火思想的改进蚁群算法。将模型和算法在某大型工程项目任务调度中加以应用,验证了所提出的优化调度方法的正确性和有效性。     

基于神经元自适应变结构控制的主动队列管理算法

针对网络TCP模型的非线性以及回路延时和负载波动等不确定性因素,提出一种基于神经元自适应变结构控制(VSC)的主动队列管理(AQM)算法。通过非线性变结构控制以保证路由器队列响应的快速性和鲁棒性;同时考虑到滑模控制中存在的抖振会引起队列波动和控制精度降低等问题,引入神经元在线调整控制器参数以减弱抖振,从而减小队列延时和模型不确定性的影响,提高AQM系统的鲁棒性和性能。最后通过NS-2仿真实验验证了算法的有效性。     

云计算环境下基于模板的任务调度策略与算法

针对云计算任务调度,提出了一种基于模板的任务调度（Template-based Task Scheduling,TTS）策略。该策略充分考虑了通信开销,在对任务分配进行预处理的基础上实现任务调度,主要分为两步：针对一个任务集合,采用可分任务调度求解子任务大小的方法,求出各个处理机应该分担的任务量模板;根据求出的模板,采用合理的调度算法对任务进行调度,从而得到较优的调度结果。在TTS策略下,对传统贪心算法加以改进,最终提出基于模板的任务调度贪心算法（Template-based Task Scheduling Greedy Algorithm,TTSGdA）。与Min-min算法和遗传算法的对比实验结果表明,TTSGdA能够有效减少任务集合完成时间。     

Cloudlet dynamic server selection policy for mobile task off-loading in mobile cloud computing using soft computing techniques

Wide acceptance of mobile phones and their resource hungry applications have highlighted resource limitations of mobile devices. In this regard, cloud computing has provided mobile phones with unlimited resources in order to help them overcome their constraints and enable them to support wider range of applications; so, mobile devices can outsource their tasks to public or local clouds. To accommodate to exponential growth of requests, user requests should be distributed to different cloudlets and then transparently and dynamically redirected to the servers according to the latest network and server status. Therefore, finding the best place to off-load is vital and crucial to both functionality and performance of the system. However, accurate and timely parameters of network and servers’ status are improbable to achieve, so the traditional algorithms cannot perform effectively and fully efficient. As a solution in this paper, an adaptive neuro-fuzzy inference system is proposed and trained to assign tasks to the servers efficiently. The trained system is robust to imprecise context information and is tolerable measurement noise and errors. We have considered improving both system performance and user quality of service parameters in this paper. Simulation results demonstrate that, compared with other server selection schemes, the proposed scheme can achieve higher resource utilization (utilization is a percentage of time that a server is busy doing something), provide better user-perceived quality of service, and efficiently deal with network dynamics. Simulation results show that our proposed algorithm excels over the compared works in terms of performance, at the best case about 30% and at the worst case about 8.93%.     

基于QoS模型感知的云作业调度算法

为了提高云作业调度的成功率,保证用户的服务质量,提出了一种基于QoS模型感知的云作业调度算法。首先从作业分为大作业和小作业两种,小作业直接在私有云平台中完成调度,然后将大作业在混合云中台上进行调度,并通过权值较小的子任务优选部署到公有云上,最后采用仿真测试实验检验算法的性能。仿真结果表明,相对于其他云作业调度算法,该算法提高了作业调度的成功率,缩短了作业的实际完工时间,获得了满足用户服务质量的云作业调度结果。     

基于粒子群优化的主动队列管理方法

针对网络拥塞现象,基于粒子群优化(PSO)提出了一种新的主动队列管理算法RQQM。该算法首先通过粒子群优化和变异算子来计算当前队列长度,并且基于到达速率和当前队列长度给出了丢包策略和丢包概率。最后,以实际数据将RQQM算法与基于速率的早期检测公平队列管理(RFED)算法和自适应主动队列管理(ABLUE)算法进行仿真实验,
发现丢包率受利用率和缓冲区影响较大;同时实验结果表明RQQM算法的公平性远远优于其他两种算法,其平均丢包率降低至12.21%。     

Energy efficient and robust allocation of interdependent tasks on mobile ad hoc computational grid

A mobile ad hoc computational grid is a distributed computing infrastructure that allows mobile nodes to share computing resources in a mobile ad hoc environment. Compared to traditional distributed systems such as grids and clouds, resource allocation in mobile ad hoc computational grids is not straightforward because of node mobility, limited battery power and an infrastructure‐less network environment. The existing schemes are either based on a decentralized architecture that results in poor allocation decisions or assume independent tasks. This paper presents a scheme that allocates interdependent tasks and aims to reduce task completion time and the amount of energy consumed in transmission of data. This scheme comprises two key algorithms: resource selection and resource allocation. The resource selection algorithm is designed to select nodes that remain connected for a longer period, whereas the resource assignment or allocation algorithm is developed to allocate interdependent tasks to the nodes that are accessible at the minimum transmission power. The scheme is based on a hybrid architecture that results in effective allocation decisions, reduces the communication cost associated with the exchange of control information, and distributes the processing burden among the nodes. The paper also investigates the relationship between the data transfer time and transmission energy consumption and presents a power‐based routing protocol to reduce data transfer costs and transmission energy consumption. Copyright © 2014 John Wiley & Sons, Ltd.     

基于时延的动态优先级调度算法

队列管理是提高网络QoS的一种有效方法。在基于时延的调度算法（BDS）基础上将时间片与优先级相结合,提出了一种基于时延的动态优先级调度算法（DDPQS）。为了实现该算法,针对进入缓冲区的每个子队列设置一个计数器,以调整的计数器值为基准来动态的改变队列的优先级,从而达到队列调度的效果;又从研究该算法的过程中,发现其局限性,即计数器值对时间片过于敏感的问题,于是进一步采用设置阈值进行区分的方法来优化。优化前后的仿真结果表明,时延和吞吐率性能具有明显改善。     

An energy-efficient task scheduling for mobile devices based on cloud assistant

Mobile cloud computing is an emerging service model to extend the capability and the battery life of mobile devices. Mostly one network application can be decomposed into fine-grained tasks which consist of sequential tasks and parallel tasks. With the assistance of mobile cloud computing, some tasks could be offloaded to the cloud for speeding up executions and saving energy. However, the task offloading results in some additional cost during the communication between cloud and mobile devices. Therefore, this paper proposes an energy-efficient scheduling of tasks, in which the mobile device offloads appropriate tasks to the cloud via a Wi-Fi access point. The scheduling aims to minimize the energy consumption of mobile device for one application under the constraint of total completion time. This task scheduling problem is reconstructed into a constrained shortest path problem and the LARAC method is applied to get the approximate optimal solution. The proposed energy-efficient strategy decreases 81.93% of energy consumption and 25.70% of time at most, compared with the local strategy. Moreover, the applicability and performance of the proposed strategy are verified in different patterns of applications, where the time constraint, the workload ratio between communication and computation are various.     

一种基于电压岛的DVFS控制算法

提出一种基于岛间队列特征的动态电压频率缩放控制算法,使用岛间队列增长率和使用率来实现电压岛工作电压/频率的动态控制。该算法引入岛间队列增长率实现了简单高效的负载预测,提高了片上通信稳定性。仿真分析表明,该算法能够更好地节能降耗。     

基于速率和队长的大时滞网络AQM算法

针对网络拥塞控制系统在大时滞网络中产生的不利影响,提出一种基于速率和队长的大时滞网络AQM算法。该算法采用缓冲区队列长度和包到达速率作为网络拥塞的判别依据,在结合Smith预估的模糊PID控制方法中加入速率控制项。仿真表明该算法在大时滞和网络动态变化的环境中拥塞响应较快、收敛时间短,并能较好地将队列长度稳定到期望值附近,提高缓冲区的利用率。     

移动云计算的任务迁移方法

移动设备在移动过程中受到网络波动的影响很大,由于工作流中任务会传输到云端执行,会带来传输与接收的能耗,再加上所处云域中微云受到不同请求数、剩余资源等因素的影响,造成了微云响应延迟、任务等待执行时间延迟和网络活跃能量损耗,再加上异构网络对数据传输造成的切换延迟影响,使任务迁移延迟更加严重,所以在上传与接收阶段采用延时传输的策略,在保证总完成时间的基础上减少传输时间;在任务迁移阶段提出了微云跳跃选择算法和网络切换算法相结合的MJSA-NHA算法,通过跳跃选择找到合适的微云以减少不必要的响应与等待延迟,并且通过切换网络空闲—活跃状态减少网络能量损耗;并在延时传输基础上加入移动速度调节算法(MSAA),在发生网络变化过程中变换移动设备的速度,这样可以在减少网络切换次数的同时减少网络切换延迟。实验结果证明,延时传输策略比Random算法在总完成时间和移动端总能耗问题上优化了很多;MJSA-NHA比MuSIC算法、task delegation和code-offloading在迁移时间上分别优化了66%~86%、36%~56%和5%~8%,在能耗方面分别优化了56%~78%、25%~46%和6%~8%;MSAA比MuSIC算法、task delegation和code-offloading在网络切换延迟上优化了50%。