Fog Computing for IoT

Services at the edge networks

Fog computing is a paradigm that provides services to users at the edge networks. The fog computing platform is typically sandwitched between the layers of the cloud servers and the users. In a traditional network, the devices at the edge layer usually perform operations related to networking such as routers, gateways, bridges, and hubs. However, in the case of the fog-enabled environment, researchers envision these devices to be capable of performing both computational and networking operations, simultaneously. Although these fog devices are resource-constrained compared to the cloud servers, the geographical spread and the decentralized nature of the fog architecture helps in offering reliable services over a wide area. Further, with fog computing, several manufacturers and service providers offer their services at affordable rates. Another advantage of fog computing is the physical location of the devices, which are closer to the users than the cloud servers, which eventually reduces operational latency significantly.

In the SWAN research group, we studied the suitability of fog computing in IoT environments and theoretically modeled its parameters to support IoT applications. We also studied its performance predominantly in IoT environments from different perspectives such as computation offloading, while reducing operational latencies and energy consumption. We also formulated models for analyzing the performance of the fog devices, while ensuring increased QoS for improved perfoamance at the user-end. Presently, we are also developing real fog-enabled platforms.

 

Assessment and Theoretical Modelling of Fog Computing

Comparative Analysis

In this work, we performed a comparative study of the fog computing paradigm with the traditional cloud architecture. Towards this, we mathematically designed the parameters affecting the devices involved, such as power consumption, latency, CO2 emission, and cost. As one of the first attempts, we assessed how fog computing helps in enhancing service provisioning for the users in IoT environments, while outperforming the cloud computing paradigm.

Read More PDF Citation

Theoretical Modelling

In this work, we presented a theoretical model for the fog computing paradigm along with definitions of the individual components responsible in the context of IoT applications. This work was one of the first attempts to provide mathematical formulations for the fog computing paradigm. We then perform a comparative study against the cloud computing paradigm and present how fog computing outperforms the former for various parameters.

Read More PDF Citation
 

Computation Offloading in Fog Computing

Autonomous Controlling of Micro-quadcopter

In this work, we proposed a method for controlling a micro-quadcopter using gestures instead of traditional remote controllers. Towards this, we designed a fog-based architecture for taking inputs through a camera and then send commands to the micro-quadcopter, which makes the flight more responsive and adaptive. The proposed architecture also provides better accuracy than the state-of-the-art technologies, such as Kinect-based hardware and CNN-based solutions. The fog-based architecture also allows stabilization of the micro-quadcopter even during the presence of unbalanced payloads, which reduces frequent replacement of micro-quadcopters.

Read More PDF Citation

SDN Topology Control

In this work, we presented an Integer Linear Program (ILP) based solution for multi-hop computation offloading in a software-defined access network. With the ILP, we make optimal decisions for local/remote computations, fog node selection, and path selection. To cope with the dynamic nature of the IoT environments, we formulate our ILP by considering essential parameters such as delay, energy consumption, multi-hop paths, and network conditions. Experimental results of this work shows reduced delay and energy consumption compared to existing solutions.

Read More PDF Citation

Devote: Criticality-Aware Federated Service Provisioning in Fog-Based IoT Environments

In this paper, we present an efficient criticality-aware decision-making system, named Devote, for fog-based Internet of Things (IoT) environment. Devote introduces an intelligent algorithm for the service of data based on the criticality, while considering the current availability of the resources at the fog node. To cope with the dynamic IoT environment, we adopt a reinforcement learning-based algorithm for the processing of the IoT data based on time-varying conditions. Additionally, we propose an efficient online secretary-based algorithm for choosingthe best suitable candidate fog node for offloading the data. To show the effectiveness of Devote, we obtained the numericalresults for assessing its performance, while collating it with the benchmark schemes.

Read More PDF Citation

PANDA: Preference-Based Bandwidth Allocation in Fog-Enabled Internet of Underground-Mine Things

In this paper, we argue that the fog computing paradigm can be leveraged to ensure safety in the Internet of Underground-Mine Things. We emphasize on prioritizing and speeding up the processing of sensitive databy introducing a factor named ‘criticality index’ and ensuring the fairness of bandwidth allocation using this index to comply to the resource requirement of critical data. We formulate the interaction between the fog nodes using the notions of the Stackelberg game. Simulation-based results show that using the proposed scheme, the fog nodes are able to allocate bandwidth with less average error and offset bandwidth, compared to the benchmark schemes.

Read More PDF Citation

Decentralized Computation Offloading

This work breakdowns the high-level tasks into smaller subtasks and form a directed acyclic task graph (DATG) to reduce operational latencies in Fog-enabled dIoT. This work proposes an ϵ -greedy nonstationary multiarmed bandit-based scheme (D2CIT) for online task allocation. The online learning D2CIT scheme allows the FN to autonomously select a set of FNs for distributing the subtasks among themselves and executes the subtasks in parallel with minimum latency, energy, and resource usage.

Read More PDF Citation

Decentralized Multiuser Computation Offloading

This work proposes a mechanism - DEFT- to facilitate mobile Internet of Vehicles (IoV) user entities (UEs) with unit tasks to cooperate among one another and offload their tasks to nearby fog nodes (FNs)/roadside units (RSUs) in a decentralized and fair manner. While existing literature depends on offloading schemes based on centralized systems, DEFT provides near real-time computation offloading in a fog-enabled IoV using a two-level game-theoretic approach. Further, this approach allows the UEs to make their own decisions in a dynamic environment.

Read More PDF Citation
 

LOAN: Latency-Aware Task Offloading in Association-Free Social Fog-IoV Networks

In this paper, we present a latency-aware task offloading scheme, named LOAN, in a fog-enabled association-free social Internet of Vehicle (IoV) network that aims to minimize the delay of time-critical tasks while saving the starvation of best-effort tasks.

Read More PDF Citation

DeTTO: Dependency-Aware Trustworthy Task Offloading in Vehicular IoT

This work investigates the dependency-aware trustworthy task offloading problem (DeTTO), especially in a fog-enabled vehicular network, where a large computation-intensive task offloaded from a vehicle is fragmented into multiple subtasks and then offloaded to multiple trusted nodes. We formulate the task offloading problem as a graph optimization problem intending to find an optimal set of trustworthy fog nodes, considered road side units (RSUs), for offloading the subtasks. Then, we minimize the task completion delay and energy consumption, while satisfying the dependency relations between the subtasks and the trust requirements of the tasks.

Read More PDF Citation

FedServ: Federated Task Service in Fog-Enabled Internet of Vehicles

We present FedServ, a federated task service system for fog-enabled Internet of vehicles (IoV), using which we aim to minimize the delay of vehicle task service. FedServ considers different task service requirements such as delay, computation-intensiveness, processing units required, and energy consumption.

Read More PDF Citation

CoTEV: Trustworthy and Cooperative Task Execution in Internet of Vehicles

In this work, we propose algorithms for cooperative task execution by taking the help of trusted vehicles, when it is not possible to complete a deadline-specified task through the RSUs. We propose a hedonic coalition formation game-based approach to form distributed coalitions of cooperative vehicles. We consider the trust score of the vehicles along with their computational capabilities and journey routes. After each task execution, the service feedback is reflected in the trust score of each cooperative vehicle in the coalition. Our proposed algorithms allow the cooperative vehicles to autonomously choose the coalitions and select a vehicle task to maximize their payoffs.

Read More PDF Citation
 

Distributed learning in Fog-enabled IoT

Fog-enabled Federated learning

This work proposes a fog-enabled federated learning framework—FogFL—to facilitate distributed learning for delay-sensitive applications in resource-constrained IoT environments. While federated learning (FL) is a popular distributed learning approach, it suffers from communication overheads and high computational requirements. Moreover, global aggregation in FL relies on a centralized server, prone to malicious attacks, resulting in inefficient training models. This work addresses these issues by introducing geospatially placed fog nodes into the FL framework as local aggregators.

Read More PDF Citation

Latency-Aware Horizontal Computation Offloading for Parallel Processing in Fog-Enabled IoT

In this paper, we propose a two-step distributed horizontal architecture for computation offloading in a fog-enabled Internet of Things (IoT) environment – HD-Fog – to minimize the overall energy consumption, and latency while executing hard real-time applications.

Read More PDF Citation

Timed Loops for Distributed Storage in Wireless Networks

In this paper, we propose the paradigm Timed Loop Storage to distribute the data and use the underutilized bandwidth of local network links for sequentially queuing packets of computational data that are being operated-on in parts in one of the IoT nodes, similar to a FIFO.

Read More PDF Citation

Magnum: A Distributed Framework for Enabling Transfer Learning in B5G-Enabled Industrial IoT

In this paper, we propose the paradigm Timed Loop Storage to distribute the data and use the underutilized bandwidth of local network links for sequentially queuing packets of computational data that are being operated-on in parts in one of the IoT nodes, similar to a FIFO.

Read More PDF Citation

Data-Centric Client Selection for Federated Learning over Distributed Edge Networks

This work presents an efficient data-centric client selection approach, named DICE, to enable federated learning (FL) over distributed edge networks. Prior research focused on assessing the computation and communication ability of the client devices for selection in FL. On-device data quality, in terms of data volume and heterogeneity, across these distributed devices is largely overlooked. The obvious outcome is the selection of an improper subset of clients with poor quality data, which inevitably results in an inefficient trained model. With an aim to address this problem, in this work, we design DICE which prioritizes the data quality of the client devices in the selection phase, in addition to their computation and communication abilities, to improve the accuracy of FL. Additionally, in DICE, we introduce the assistance of vicinal edge devices to account for the lack of computation or communication abilities in certain devices without violating the privacy-preserving guarantees of FL. Towards this aim, we propose a scheme to decide the optimal edge device, in terms of latency and workload, to be selected as the helper device. The experimental results show that DICE improves convergence speed for a given level of model accuracy.

Read More PDF Citation