IoT Applications on 5G Edge

IoT Applications on 5G Edge

Shelley Sam Varughese
Student: Department of Computer Science
Musaliar College of Engineering and Technology,
Affiliated to APJ Abdul Kalam Technological University,
Pathanamthitta, Kerala, India.

Prof. Deepa Thomas
Assistant Professor: Department of Computer Science
Musaliar College of Engineering and Technology,
Affiliated to APJ Abdul Kalam Technological University,
Pathanamthitta, Kerala, India.

Dr. Manikandan L C
Head of the Department: Department of Computer Science
Musaliar College of Engineering and Technology,
Affiliated to APJ Abdul Kalam Technological University,
Pathanamthitta, Kerala, India.

Abstract: Internet of things (IOT) primarily consists of a system that connects to the internet. The IoT related concepts like self-driving cars, smart cities, e-health care, etc. have a ubiquitous presence now. These applications require higher data-rates, larger bandwidth, increased capacity, low latency and high throughput. The key for the future IoT system is the shift from a static architecture to a dynamically self-organizing and evolving one. Devices of extremely varying capabilities needs to collaborate and access all necessary information to ensure their optimal work, while keeping the flexibility in the network configuration. The supporting infrastructure should allow the connected devices to interact with the most convenient node in the network and should be able to optimize the resource consumption, without compromising QoS. The main idea is to summarize the process of building an offloading framework for the arbitrary task of IoT devices. 5G cellular networks provide the enabling technologies for deployment of the IoT technology everywhere and anywhere. Another contribution is the collection of requirements like need for Edge Computing, Offloading paradigms and an underlying infrastructure for the IoT networks and computing systems, which will be supported by means of Clustering, 5G mobile networking standards, and Artificial Intelligence.

Keywords IoT networks, Edge Computing,5G, Artificial Intelligence.

I. INTRODUCTION

IoT applications handles informations from a number of heterogeneous devices. 5g is the foundation for realizing full potential of IoT. In the past generation IoT devices which were considered as small, external hardwares with limited resources, like sensors which resides at the edge of the involved network infrastructure. This assumption made the main role of the IoT devices to blindly transmit sensed data or to react to environmental changes to some extent.

Fig 1: Block diagram showing IoT based smart applications

The limitation on the computing resources on the IoT devices, made practice is of offloading tasks of various applications to the computing systems with resources like data centres in the cloud. However, the main drawbacks of the offloading methods were high latency and network

congestion in the infrastructure. This issue gave rise to the paradigm of Edge Computing, the idea was to support the devices with a cloud closer to the edge of the network. This appeared as a solution. However, adding Edge resources complicated the management of the network because multiple devices will be contending them.

Furthermore, the recent evolution of IoT brought more and more devices which were not simple sensors or transmitters. It provided a limited execution environment. This opened up a huge opportunity to utilize this previously unused processing power in order to offload custom application logic directly to these edge devices. In this very complex scenario, it is an essential question of how to balance the tasks and the resources available in such a way that would profit from the added capabilities of the IoT devices without compromising the final performance.

The key question in future of IoT networks then was how to enable devices of extremely varying capabilities to collaborate and access all information necessary for ensuring the optimal work while keeping the flexibility in the network configuration. An important issue in relation with collaboration of IoT devices is connectivity, since the continuously growing number of devices could generate congestions in the communication channels.

The last change in the IoT is mobility. We would have to consider many kinds of objects. The definition of Things is very broad, consisting from smart phones to even smart cars. Things are actually any physical objects which have a real- life presence. Such objects could be installed on moving vehicles will be mobile themselves.

For all presented scenarios, there will be dynamic changes in configuration at the edge of the network. This happens very frequently. For example, the connected devices may be physically moving and the network might need to balance the resources or will have to reallocate them to achieve system faults tolerance. These continuous changes will require the IoT networks to be able to reorganize itself in such a way that optimizes the resource consumption like bandwidth, storage and power. And allows the connected devices to interact with the most convenient node in the network, without compromising the Quality of the Service (QoS). Simply copying the whole applications in every node which requires them cannot be scalable or a maintainable solution and offloading framework is still needed.

The answer to the connectivity question of the ever- increasing number of devices is a solution which is already applied for Wireless Sensor Networks (WSNs) and it is to form groups of devices and to manage its connections in a collaborative way.

A network of such devices could profit from the 5th Generation (5G) mobile network standards and its infrastructures in order to achieve the goals which is

otherwise unattainable with only Edge computing and offloading.

The network should be able to adapt to fluctuations of resource load. As the 5G networks are more complex architectures than their predecessors, the number of configuration variable makes it very difficult to apply the deterministic adjustment approaches. Due to this, we could believe that the IoT framework will also benefit from the Artificial Intelligence (AI) based technologies which are designed precisely to cope with these challenges.

The main effort is to summarize and redefine the requirements for this new IoT networks and computing systems, the need for a platform on top of 5G edge computing and computational offloading paradigms. Compared to previous works we have go into more details regarding the possible technical solution that can be used to satisfy the requirements of this IoT platform, in particular we could give the overview of the different choices for modelling and partitioning the application and we could individuate some progress in clustering and AI that can support the development of such complex solutions.

In the effort to summarize the processes needed for building an offloading framework for arbitrary task of the IoT devices we specified five main parts:

Discovery and modelling

Planning and optimization

Execution

Monitoring and performance maintenance

Learning and predicting

II. BACKGROUND

Fig 2: Block diagram showing IoT benefits as the cutting-edge technology In this section the basic definitions of the involved technologies and paradigms are tried to summarize the mainstream directions of the literature.

A. EDGE COMPUTING

As the number of mobile devices and services that requires computational or storage capabilities that significantly exceeded their own capacities, the paradigm of Cloud Computing (CC) rose and gained a continuously increasing importance. The concept of cloud computing is based on Data Centres (DC), which are capable of managing the processing and storage requirements of the tasks involving very large data. Moreover, Data Centre Networks (DCNs) are built up by connecting the data centres using optical cables. Due to the extremely low internal

communication costs it is appeared as a single entity inthe outside world. When a problem occurs that will outrun the available resources one can offload the data and the code to the cloud and after the computations are done the results are given back.

The paradigm of Cloud computing has thus given a solution to the scalability related problems due to inefficient resources. There is a significant latency or congestion in the network due to the code and data migration. Location unawareness CC paradigm is the cause of the problem and it will get more severe time as more and more (semi-)intelligent devices attempts to connect to DCNs.

The concept of Edge Computing emerged and it leveraged the storage and computation capabilities of the edge devices which are connected to the Internet and was meant to be an intermediate layer between the devices. They are able to handle the subset of requests is usually sent to the cloud, but didnt need its real involvement due to the diminished resource requirement or because of the no need for the DC to be involved. Thus, the computation load of DCs is reduced to some extent. The latency of responses when an application needs to have real-time or almost real-time responses also have been reduced. Moreover, due to its high availability and geo-distributed nature, the Edge layer is appropriate to handle the challenges of mobility; for e.g., serve the requests of moving users like in the cases of autonomous cars or streaming and real-time gaming.

1) Comparison of Edge Computing Implementations: Fog Computing, Cloudlet and Mobile Edge Computing: In the following the distinction of Edge Computing into three categories.

a) Fog Computing: In this, Fog Computing Nodes can be placed at any point of the architecture. They are highly heterogeneous and can be built on various devices such as routers, switches, IoT gateways etc. Heterogeneity of devices leaded to the ability for working with different protocols and with non-IP- based technologies that is in communication between fog computing nodes and the end devices. Since heterogeneity of the edge naturally stays hidden from the user devices. Fog computing system exposes a uniform interface containing storage and computational services. Also, the monitoring security and device management facilities are exposed. On top of this abstraction layer, the orchestration layer organizes resource allocations according to the users requests.

b) Cloudlet: Cloudlet can be defined as a trusted set of computers which have a good connection to the Internet and their resources are made available to nearby mobile devices. The Cloudlet runs a virtual machine which is capable of arranging resources to the connected users near real-time over the WLAN network within one-hop distance and high bandwidth. Above provisioning the infrastructure, Cloudlet architecture provides a middle framework support to component-based applications designed

B. 5GC. OFFLOADING AND PARTITIONINGD. CLUSTERING FOR WSN AND IOT SYSTEMSE. AI IN NETWORKINGA. General IoT RequirementsB. RELATED WORKS ON IOT INFRASTRUCTURESA. 5G EDGE FOR IOTB. CLUSTERING FOR IOT1) CHALLANGES IN CLUSTERING FOR 5G ANDC. AI IN 5G

Next >