The concept of the Internet of Everything (IoE) is an area of great interest, with a lot of attention. The fact that it is used in almost all areas makes this concept progress much faster. The devices used in the envi...
详细信息
ISBN:
(纸本)9781728112442
The concept of the Internet of Everything (IoE) is an area of great interest, with a lot of attention. The fact that it is used in almost all areas makes this concept progress much faster. The devices used in the environments of this concept are required to be minimal in terms of hardware as well as their energy consumption. Although there are different ways of minimizing energy consumption in these devices, one of the most promising protocols that provide mentioned demands is the IEEE 802.15.4e Time Slotted Channel Hopping (TSCH) Medium Access Control (MAC) protocol. This protocol runs on 802.15.4 radio and requires a communication schedule for MAC. How and when these schedules, updates and maintenance will take place are not covered by this protocol. Several solutions are available in the literature to solve these problems. Distributed divergecast Aloha based scheduling and DIVA scheduling algorithms are two of these solutions. Both algorithms make their scheduling according to certain probability calculations. Changing the values of these probability calculations is a novel research with evaluation results in this paper. The aim of this study is to examine the effect of the change in the related probability variables on the performance and try to find the best possible values. It has been observed that the changes in these calculation values have an effect on network converge time.
In recent years, Artificial Neural Networks have evolved rapidly and are applied to various fields. Meanwhile, to enhance computation efficiency of neural network applications, more and more neural network accelerator...
详细信息
ISBN:
(纸本)9783030296117;9783030296100
In recent years, Artificial Neural Networks have evolved rapidly and are applied to various fields. Meanwhile, to enhance computation efficiency of neural network applications, more and more neural network accelerators have been developed. Though traditional task scheduling algorithms on heterogeneous systems have been intensively researched, they can't be applied to neural network accelerators directly. Based on typical characteristics of neural network accelerators, we formalize the problem of tasks scheduling for neural networks, and transplant two listing heuristic scheduling algorithms, Heterogeneous-Earliest-Finish-Time (HEFT) and Critical-Pathon-a-Processor (CPOP). Inspired by the separable features of neural network operations, we propose two partition algorithms, the Iterative Partition scheduling Algorithm (IPS) and the Partition scheduling Combination Algorithm (PSC), which can be associated with scheduling algorithms. Further, we conduct experiments on some typical neural networks, and results show that compared to scheduling-only algorithms the partition associated algorithms achieve about 2x to 3x speedup.
With dramatically increasing demand for data usage in the LTE Heterogeneous Networks (HetNets), traffic offloading techniques have been used to balance the load among the network tiers in order to enhance the overall ...
详细信息
ISBN:
(纸本)9781728107295
With dramatically increasing demand for data usage in the LTE Heterogeneous Networks (HetNets), traffic offloading techniques have been used to balance the load among the network tiers in order to enhance the overall system performance. The Cell Range Expansion (CRE) and Almost Blank Subframe (ABS) are the two major techniques under the enhanced Inter-Cell Interference Coordination (eICIC), which are aimed at increasing overall system performance while keeping low interference toward cell edge users. In general, a scheduling algorithm is also one major issue affecting the system efficiency. In this paper, different scheduling schemes, i.e. Round Robin (RR), Proportional Fairness (PF), Best-Channel Quality Identification (Best-CQI), Maximum Throughput (Max-TP) and Resource Fairness (RF), are investigated. The study focuses on the comparison of the system performance when operating with different schedulers as well as under different CRE and ABS parameters. The results are observed through the system level simulation in terms of average throughput, peak throughput, edge throughput, and fairness.
In cloud computing, resources are provided as a public utility and the user can lease and release those resources via the Internet by an on-demand fashion. Since cloud resources are meant to be utilized properly, the ...
详细信息
ISBN:
(纸本)9781538681251
In cloud computing, resources are provided as a public utility and the user can lease and release those resources via the Internet by an on-demand fashion. Since cloud resources are meant to be utilized properly, the role of scheduling algorithms is vital to ensure an appropriate resource is available to every request. In this paper, major cloud computing scheduling algorithms are reviewed, and discussion is made accordingly. This paper also compares cloud computing scheduling from the perspective of makespan, load balancing, CPU utilization, deadline, response time, and allocation cost. In addition, the paper proposes an abstract model to integrate desirable features the of algorithm suitable to cloud environment. Future research opportunities are highlighted and the end of the paper.
Hadoop bundles the two computing resources of memory and CPU in the management resources, and then divides it into two resource models: MapSlot and ReduceSlot according to task types. MapReduce applications will have ...
详细信息
Hadoop bundles the two computing resources of memory and CPU in the management resources, and then divides it into two resource models: MapSlot and ReduceSlot according to task types. MapReduce applications will have a large number of sorting operations in operation. Most of these sorts are executed iteratively, which consumes a lot of performance. Chapter 5 of this article takes this as an entry point and reorganizes the execution process of the Shuffle stage. Researched to replace quick sort with more efficient counting sorting. At the same time, the Shuffle execution is branched according to the definition of Combiner. One branch deletes the quick sort in the partition in the spill phase and the merge sort in the combine phase to reduce performance consumption. The other branch executes Combiner in advance to improve data processing efficiency. The two branches processed 21GB of log data on a 7-node PC cluster, and both achieved an efficiency improvement of about half an hour.
This paper provides a survey of the state-of-the-art workflow scheduling algorithms with the assumption of cloud computing being used as the underlying compute infrastructure in support of large-scale scientific workf...
详细信息
ISBN:
(数字)9781728187891
ISBN:
(纸本)9781728187907
This paper provides a survey of the state-of-the-art workflow scheduling algorithms with the assumption of cloud computing being used as the underlying compute infrastructure in support of large-scale scientific workflows involving big data. The survey also reviews a few selected representative scientific workflow systems in light of usability, performance, popularity, and other prominent features. In contrast to existing related surveys, which most try to be comprehensive in coverage and inevitably fall short in the depth of their coverage on workflow scheduling, this survey puts an emphasis on the two dominant factors in workflow scheduling, the makespan and the monetary cost of workflow execution, resulted in a useful taxonomy of workflow scheduling algorithms as an additional contribution. This survey tries to maintain a good balance between width and depth in its coverage - after a broad review, it spotlights on selected top ten representative scheduling algorithms and top five workflow management systems leveraging cloud infrastructure with an emphasis on support for big data scientific workflows.
This work provides some criteria that allow network administrators to measure the impact of different network configurations have on voice, data, and video applications quality of service offered to the different user...
详细信息
ISBN:
(数字)9781728173498
ISBN:
(纸本)9781728173504
This work provides some criteria that allow network administrators to measure the impact of different network configurations have on voice, data, and video applications quality of service offered to the different users. For that, this project implemented tests to measure parameters such as packet loss, delay and jitter using static routing, different packet scheduling algorithms (Custom Queuing, Priority Queuing, and Weighted First Queuing), different services (voice, data, and video), and connection speeds using Cisco 2800 Routers and D-ITG traffic generator in conjunction with the Network Time Protocol synchronizer. A comparative analysis of sensitive parameters for Quality of Service was conducted to describe network behavior for each service. The implemented services presented performances according to quality requirements, reported in the literature. Specifically, the data service presented delay and jitter within the levels considered acceptable for this application and zero packet loss. For its part, the video service presented levels of delay and jitter according to the quality requirements for streaming. Finally, the voice service presented the best configuration performance with the Priority Queuing algorithm, for all measured service quality parameters.
Motivated by applications in data center networks, in this paper, we study the problem of scheduling in an input queued switch. While throughput maximizing algorithms in a switch are well-understood, delay analysis wa...
详细信息
We propose a way to model and integrate HPC scheduling simulators into a popular Reinforcement Learning toolkit. We show experimentally that such an approach not only aids researchers being able to iterate faster by m...
详细信息
ISBN:
(数字)9781728192383
ISBN:
(纸本)9781728192390
We propose a way to model and integrate HPC scheduling simulators into a popular Reinforcement Learning toolkit. We show experimentally that such an approach not only aids researchers being able to iterate faster by means of software reuse, but also to achieve state-of-the-art performance with 10x less interactions with the environment. We validate the simulation model's correctness by using unit tests, assertions and experimental comparisons. We also share an open source implementation of the model that will benefit researchers in resource management tasks assisted by Machine Learning.
Traffic lights have and always will be necessary for the safety of the traffic on the road. However, the time cycles of these traffic lights are based on premeditated computations and not real time data. These schedul...
详细信息
暂无评论