Edge computing

About: Edge computing is a research topic. Over the lifetime, 11657 publications have been published within this topic receiving 148533 citations.

Prioritized task scheduling in fog computing

Abstract: Fog computing, similar to edge computing, has been proposed as a model to introduce a virtualized layer between the end users and the back-end cloud data centers. Fog computing has attracted much attention due to the recent rapid deployment of smart devices and Internet-of-Things (IoT) systems, which often requires real-time, stringent-delay services. The fog layer placed between client and cloud layers aims to reduce the delay in terms of transmission and processing times, as well as the overall cost. To support the increasing number of IoT, smart devices, and to improve performance and reduce cost, this paper proposes a task scheduling algorithm in the fog layer based on priority levels. The proposed architecture, queueing and priority models, priority assignment module, and the priority-based task scheduling algorithms are carefully described. Performance evaluation shows that, comparing with existing task scheduling algorithms, the proposed algorithm reduces the overall response time and notably decreases the total cost. We believe that this work is significant to the emerging fog computing technology, and the priority-based algorithm is useful to a wide range of application domains.

A hybrid secure routing and monitoring mechanism in IoT-based wireless sensor networks

Abstract: Internet of Things (IoT) has advanced its pervasiveness across the globe for the development of smart networks. It is aimed to deploy network edge that enables smart services and computation for the IoT devices. In addition, this deployment would not only improve the user experience but also provide service resiliency in case of any catastrophes. In IoT applications, the edge computing exploits distributed architecture and closeness of end-users to provide faster response and better quality of service. However, the security concern is majorly addressed to resist the vulnerability of attacks (VoA). Existing methodologies deal only with static wireless sensor web to deduce the intrusions in which the sensor nodes are deployed in a uniform manner to retain the constancy. Since the sensor nodes are constantly being in question through different transmission regions with several levels of velocities, selection of sensor monitoring nodes or guard nodes has become a challenging job in recent research. In addition, the adversaries are also moving from one location to another to explore its specific chores around the network. Thus, to provide flexible security, we propose a secure routing and monitoring protocol with multi-variant tuples using Two-Fish (TF) symmetric key approach to discover and prevent the adversaries in the global sensor network. The proposed approach is designed on the basis of the Authentication and Encryption Model (ATE). Using Eligibility Weight Function (EWF), the sensor guard nodes are selected and it is hidden with the help of complex symmetric key approach. A secure hybrid routing protocol is chosen to be built by inheriting the properties of both Multipath Optimized Link State Routing (OLSR) and Ad hoc On-Demand Multipath Distance Vector (AOMDV) protocols. The result of the proposed approach is shown that it has a high percentage of monitoring nodes in comparison with the existing routing schemes. Moreover, the proposed routing mechanism is resilient to multiple mobile adversaries; and hence it ensures multipath delivery.

BottleNet++: An End-to-End Approach for Feature Compression in Device-Edge Co-Inference Systems

Abstract: The emergence of various intelligent mobile applications demands the deployment of powerful deep learning models at resource-constrained mobile devices. The device-edge co-inference framework provides a promising solution by splitting a neural network at a mobile device and an edge computing server. In order to balance the on-device computation and the communication overhead, the splitting point needs to be carefully picked, while the intermediate feature needs to be compressed before transmission. Existing studies decoupled the design of model splitting, feature compression, and communication, which may lead to excessive resource consumption of the mobile device. In this paper, we introduce an end-to-end architecture, named BottleNet++, that consists of an encoder, a non-trainable channel layer, and a decoder for more efficient feature compression and transmission. The encoder and decoder essentially implement joint source-channel coding via lightweight convolutional neural networks (CNNs), while explicitly considering the effect of channel noise. By exploiting the strong sparsity and the fault-tolerant property of the intermediate feature in deep neural network (DNNs), BottleNet++ achieves a much higher compression ratio than existing methods. Compared with merely transmitting intermediate data without feature compression, BottleNet++ achieves up to 64× bandwidth reduction over the additive white Gaussian noise channel and up to 256× bit compression ratio in the binary erasure channel, with less than 2% reduction in accuracy of classification.