EDGE COMPUTING: CHALLENGES FOR REAL TIME ANALYTICS

Abstract

The devices that are connected all time usually withdraw benefits from cloud. IoT manufacturers however had started discovering benefits of on-device analytics in real time. This is where edge computing comes into light. Edge computing is a latest approach of handling and analyzing data generated by IoT devices for critical applications in real time. It also helps in reducing latency and dependency of such applications on cloud. Edge is considered as the counterpart of cloud where processing capabilities are restricted to the edge devices, gateways, routers etc. To harness the potential of edge computing real time analytics has been introduced at the edge. Real time analytics at edge is a boon for many critical applications especially health care where real time data processing is helping in saving lives of people. Though there are many benefits of implementing real time analytics at edge but there are many challenging issues for real time analytics at edge. The paper will discuss various challenges that lies while implementing real time analytics ate edge and also the preliminary solution to handle it.

Headnote

Abstract: The devices that are connected all time usually withdraw benefits from cloud. IoT manufacturers however had started discovering benefits of on-device analytics in real time. This is where edge computing comes into light. Edge computing is a latest approach of handling and analyzing data generated by IoT devices for critical applications in real time. It also helps in reducing latency and dependency of such applications on cloud. Edge is considered as the counterpart of cloud where processing capabilities are restricted to the edge devices, gateways, routers etc. To harness the potential of edge computing real time analytics has been introduced at the edge. Real time analytics at edge is a boon for many critical applications especially health care where real time data processing is helping in saving lives of people. Though there are many benefits of implementing real time analytics at edge but there are many challenging issues for real time analytics at edge. The paper will discuss various challenges that lies while implementing real time analytics ate edge and also the preliminary solution to handle it.

Keywords: IoT, Edge computing, Real Time Analytics, Architecture of Edge Computing

INTRODUCTION

IoT has brought the revolution in the way each domain of life is working whether it is about health care sector, education, business, security or any other stream. According to Statista, the installed base of IoT devices is forecasted to grow to almost 31 billion worldwide.

It is expected by Cisco Systems that by the end of 2021 number of IoT devices are going to reach 13.7 billion. These will generate huge demand of data centers and central cloud resources. IoT along with cloud computing is considered as a solution for managing processing, extraction and aggregation of such huge volume of data.

Due to such huge volume of growing number of IoT devices cloud computing is the solution to deal with such enormous amount of data which is generated by these devices worldwide. IoT and cloud computing are intertwined as they go hand in hand. In other words we can say that IoT devices depend largely on cloud computing for processing and retrieval of data. The traditional IoT architecture collects data from sensors, acutators, wearables and various others dispersed IoT devices. The collected data is being transmitted to the central repository where it is processed collectively or transmitted further. [ 1 ]

In the last several years cloud has been used popularly in IoT setup for centralized computing services in order to exploit the benefits of shared data center. There are many factors which have a great impact on the performance of IoT devices in cloud environment. Latency, Packet delays, virtualization and placement of data server center are major issues in cloud computing environment. Due to these issues of cloud computing Edge computing enters the scenario. Processing data which is quite close to the target devices is called Edge computing.

CLOUD COMPUTING TO EDGE COMPUTING

The coming decade is ready to witness the transformation of cloud computing to edge computing. Edge computing refers to the processing at edge rather than processing at central location like in cloud computing. Edge computing processes the data near to the data source. Edge computing performs the computational task near to the target devices. The unlimited growth of data in IoT setup and limitations of network layer are leading to the need of edge computing scenario. Edge computing processes the data physically close to the target devices, or on the device itself.

The major motive of edge computing is to deliver concrete data or information to the end users in many of the use cases especially healthcare. It is considered as a great help in reducing cost of connectivity by simply transferring data or information that is important instead of sending bulk of raw data produced from various IoT sensors. Besides that filtering data and sending only crucial data to another end reduces load on network and decreases the need of computing resources.

Edge computing is considered as an improvement over cloud. It reduces the amount of data that has to be carried to the cloud for processing and analysis purpose. The major benefit of Edge computing is low latency which is quite high in comparison of cloud. Edge networking offers resource optimization over cloud computing system. All computations are performed at the edge of the network which helps in reducing network traffic hence decreased the risk of a data bottleneck. [2]

BENEFITS OF EDGE COMPUTING

The complete scenario of data processing is changed in the last decades. The credit goes to the miniaturization of processing and storage technology. IoT devices are getting more powerful day by day in order of storage and processing of data. The doors of opportunities for the organizations have opened up to shift processing functions closer to the data source inspite of processing it on central location. Processing of such data can be performed in real time which is near to the end user.

In the edge scenario data need not to travel to the central repository, edge computing can reduce the latency and can improve performance. The efficiency and flexibility offered by edge approach to handle data creates versatile possibilities for the organization. The major benefits that can be drawn by edge computing are as follows=

= Hardware

In today's scenario requirement of strong and specialized hardware is there. Voluminous data with modern algorithms like machine learning rely on specialized hardware like TPU (Tensor Processing Units). In general for such specialized hardware scenario edge nodes are preferable.

= Latency

In comparison to cloud edge computing provides improved latency. Many applications rely on instant feedback in order to make real time decisions like - health care applications. In such cases sending data to the cloud, processing it and sending it back again may take too long. However if the data is processed at the edge only the real time decision making in such applications can be realized.

= Data Throughput

IoT devices may produce voluminous amount of data. For example- one autonomous car produces up to 4000 gigabytes of data per day. Imagine the volume of data that reaches to the central repository produced by such cars. This huge volume of data creates a load on the network. In edge computing necessary data can be processed on the edge nodes close to the devices hence reducing the load on the network.

= Reliable

Sometimes establishing connection with the central cloud is not reliable as data connectivity is poor at remote locations. Shifting the processing nodes at edge ensures reliability factor of data processing. Besides that in case if one edge device fails another edge will take over.

= Privacy

There are many IoT use cases where data is crucial and is required to be collected for further processing. For example - In healthcare system patient's data is collected for further analysis and processing. Preserving privacy of such data is quite crucial; hence keeping it on edge is safer than cloud.

= Scalability

Upgradation of cloud infrastructure as per the new use case requirement is not an easy task and hence limits the use case adoption rate. Edge however does not suffer from such problems and can be extended easily. Extension of edge computing framework is quite easy and automated process by simply adding, replacing or upgrading an edge node.

ARCHITECTURE OF EDGE COMPUTING

The boundaries of the edge computing cannot be defined. It varies application to application. In fact its boundaries are logical rather than physical. The overall Edge computing architecture comprises of following components:

= Cloud layer: This layer seems to be costly in terms of data transport and latency although the computational power and storage resources are endless. In edge computing environment the role of cloud can be as storage area and critical processing unit of complex data.

= Edge Node: These are powerful nodes that are located at the endpoint of the network. These nodes are devices that are capable of network traffic and do possess high computational power. The edge nodes can be switches, routers or sometimes it can be small-scale data centers as well.

= Edge Gateway: It works for processing of data from an edge device and after processing only relevant data that is required for further processing is transported to cloud in order to reduce network traffic. Sometimes that data is sent back to the edge device itself in case of applications where real-time data is processed and some analytical results are being used.

= Edge Devices: The layer consists of various devices and sensors like wearables, traffic lights, environmental sensors with limited processing power. The processing capabilities of these devices are limited. [3]

EDGE COMPUTING AND REAL TIME ANALYTICS

The basic concept of edge computing is to perform processing and analytics at the endpoint of device itself or somewhere in the edge network itself besides sending that data to the central cloud. There are numerous IoT and edge computing devices which are scattered in abundance throughout the geographically dispersed computing infrastructure. The majority of these computing resources are underutilized.

The data generated by the various devices at the age loosed its importance as it is not able to reach to the cloud due to the various issues like cost of networking, latency and communication between edge devices. These limitations are because cloud models are not supporting analytics at the huge volume of data produced by the sensors which require high latency and response time [4]

Cloud centers had huge storage and computing facilities which is easily available and it is easier to offload much of the computation tasks to the cloud. However the performance is sometimes degraded in the real time applications due to the latency issue. The processes that are delay sensitive need to be processed at edge to ensure processing of data in a real time and provide feedback to the end-users.

Edge computing has been a revolutionary change in the way data is being processed and handled which is produced from millions of devices scattered around the world. The growth of IoT devices and requirement of real-time data processing applications is the major driver of demand of edge computing systems. The latest, faster wireless networking technologies like 5G are the promotional factor for the creation and application of real-time systems such as health care, video data processing and analysis, automated cars to name a few.

The major motive of implementation of edge computing at initial stage was to handle the issue of latency. With the explosive growth of IoT devices huge volume of real-time data is generated which needs to be processed in real time hence the driving the edge technology. The edge computing is capable of processing and storing data faster hence enabling more efficient real-time applications that are critical to companies. Earlier in the applications where persons face is scanned in smartphones, the algorithm to recognize face is executed through a cloud service, which takes ample of time to process it. Edge computing can resolve the time delay by processing the algorithm on the edge server or gateway or may be in the smartphone itself.

-Use Cases of Edge with Real Time Analysis

Following are the use cases where edge computing technology can be exploited well:

= The autonomous vehicles will perform well in edge computing scenario. An autonomous vehicle does not go well with cloud. Edge computing processes the data close to the data where it is produced. Autonomous vehicles depend on short and predictable response times. The decision to make autonomous vehicles stop at pedestrian crossing should be handled immediately. Relying on remote computers for such time of reasoning will be hazardous. In such cases real time processing of the data will be done on the vehicle that minimizes the network overhead. The data needs to be processed further should be transferred to the cloud. Using edge computing can be an efficient way to communicate and send data or information regarding accidents, traffic status or weather to the central remote server.

= The healthcare organizations have to perform brainstorming in order to identify which service will have benefit from edge service, like EHR systems, telemedicine, digital imaging to name a few. Organizations have to assess each and every aspect like network bandwidth to handle needs of remote patients and what other devices need to be connected in near future. The wearable devices keep an eye on chronic conditions for patients. The edge computing with real time analysis may work instantly as a life saver by alarming caregivers whenever help is required.

= Security is another area of application of real-time edge computing. Surveillance systems will be benefitted by edge computing technology. The potential threats can be identified in the real time and will alert user by performing real time analysis.

= Smart Speakers have the ability to interpret the instructions based on voice to run basic commands. Switching on/off lights, maintaining temperature settings, saving power all such issues can be handled with the help of real time edge computing.

= Retail Advertising: In the retail advertising various demographic information of the customer is tracked and stored on cloud for further analysis. Edge computing can ensure the privacy of data by keeping it in encrypted form near to the source rather than sending that data to the central cloud in unprotected manner.

= Video Conferencing: Edge can be solution for the poor quality video streaming down from the cloud. By placing the server-side of video conferencing software closer to participants, quality problems can be reduced.

-Challenges of Real Time Analytics in Edge Computing

The IoT technologies have a great impact on the quality of life of the citizens. The IoT devices had felt its presence in every domain of our lives. Few of the application areas of IoT devices are healthcare, smart city, public lights and water supply networks. Huge volume of such data is being generated by all such setups. To handle such huge volume of data, the services of cloud is been taken for storage and processing of such data. However processing of data at cloud may face some problems of data transmission such as the interruption of the network and the mobility of users. Hence the need of edge computing is there. Edge computing is quite promising and it provides real-time data processing and analysis mechanism. [4]

Besides considering fusion of edge with Big Data and IoT paradigms, there is a need of considering fusion of cloud with edge computing. The collaboration will provide the understanding the benefits of fusion of edge resources with cloud and Big data in terms of services and communication facilities. [6-9]. Though the edge computing seems to be promising for RealTime Analytics it faces certain challenges=

= computing architecture

Edge computing is still in infancy stage and hence there is a lack of standard architecture. This is one of the challenging issues. Dealing with data generated from heterogeneous source and processing it in real time in order to draw benefits in majority of application is the major motive of Real Time analytics of edge computing. Unavailability of standard/predefined architecture puts a challenge for the implementation of edge computing scenario.

Solution to the problem is to predefine the architecture of edge while implementing in any enterprise or application. Properly defined architecture will help in definition of the data processing boundaries that which set of data will be processed at edge node and where the result will be taken for further analysis.

= Limited processing power

In comparison of cloud computing which have huge processing capacity, edge computing has limited processing capacity to perform real-time analytics sometimes complex algorithms are required which need sufficient processing power due to the limited available size

The solution to the mentioned challenge is to combine cloud and edge processing capabilities in order to give quality results. For example Lone Star Analysis (LSA) the company which is working in association with Accenture to deliver edge analytics is using combined processing capabilities of edge and cloud. It performs complex analytics on a footprint to perform predictive analysis at edge and for delayed processing it work in conjunction with cloud. [ 10]

= Positioning of Real Time Analytics

Implementation of real-time analytics by the enterprises involves various important activities like requirement gathering, design of solution architecture and also choosing the correct technology stack. Focusing on such technical tasks enterprises fail to pay attention on internal processes. The enterprises need to focus on their process of working and where to position real time analytics to withdraw maximum benefit of it. For example, A manufacturing company wants to improve their equipment repair time. Breakdowns may occur unexpectedly and the maintenance team is unable to find out cause of breakdown in a desired time, and sometimes they do not have a particular part needs to be replaced. Hence only identification of a problem in real time will not give you promising results. Organizations have to take care about their internal processes too. The people need to understand that only placing real time analytics at right position is not only the solution.

The solution of the challenge is to coordinate internal processes of the enterprise with real time analytics. Both will work in coordination and will provide promising results.

= Collection of Data

The collection of relevant data is a critical step in real time analytics application. Any error at the time of data collection will lead to the propagation of errors in application and it also affects the integrity of any application. The real time analytics rely on relevant data - taking a use case of health care systems where collection of irrelevant or incorrect data will lead to many causalities.

The solution to the mentioned issue can be filtering and verification of data while moving out and entering the premises of real time analytics application.

= Analysis of Data

Another challenging issue is fulfilling the demand of the end user by analyzing data in a real-time. Taking an example of Spotify, which is one of the popular music streaming service which saves the used playlist and next song is played automatically. However many end users do not find it exciting as they wish to listen to the music depending upon various factors like - time, place and weather. Giving such kind of service will justify the role of real-time analytics.

Spotify came up with the solution of this problem by using deep-level contextual analysis in order to justify the role of real time analytics in order to strengthen the customer satisfaction. [11]

CONCLUSION

Edge computing is providing better solutions to the companies for decision making in real time. The basic advantage of edge computing is to process the data near to its source. Processing it near to the source reduces latency and increases the efficiency in terms of time. There are many IoT applications which can perform outstanding if they are supported by the benefits of edge computing with real time analytics. Implementing real time analytics at edge faces few challenges at discussed in the paper. All such challenges if taken care properly with the mentioned solutions can lead to the successful implementation of real time analytics at edge. Successful implementation of real time analytics at edge will definitely lead to the promising outputs in various IoT applications.

References

References

[1] PS Divya, An Introduction to Edge Computing, https://dzone.com/articles/an-introduction-toedge-computing.June, 2019

[2] T Raj, Why Edge computing is critical for IoT, https://www.networkworld.com/article/3234708/ why-edge-computing-is-critical-for-the-iot.html October, 2017

[3] Maximilian Bischoff, Johannes M. Scheuermann, Christoph Kiesl und Julian Hatzky, The Edge is near: An introduction to Edge Computing, https://www.inovex.de/blog/edgecomputing-introduction/,2019

[4] Nastic S, Rausch T, Scekic O, and Dustdar S, "A Serverless Real-Time Data Analytics Platform for Edge Computing",IEEE, Internet Computing, Volume: 21 , Issue: 4,2017

[5] Bangui H, Rakrak S, Raghay S, Buhnova B, Moving to the Edge-Cloud-of Things: Recent Advances and Future Research Directions, Electronics, 2018

[6] Sookhak, M.; Yu, F.R.; Zomaya, A.Y. Auditing Big Data Storage in Cloud Computing Using Divide and Conquer Tables. IEEE Trans. Parallel Distrib. Syst. 2018, 29, 999-1012.

[7] Wan, J.; Tang, S.; Hua, Q.; Li, D.; Liu, C.; Lloret, J. Context-aware cloud robotics for material handling in cognitive industrial internet of things. IEEE Internet Things J. 2018, 5, 2272-2281.

[8] Jayasena, K.P.N.; Li, L.; Xie, Q. Multi-modal multimedia big data analyzing architecture and resource allocation on cloud platform. Neurocomputing 2017, 253, 135-143.

[9] Chard, R.; Chard, K.; Wolski, R.; Madduri, R.; Ng, B.; Bubendorfer, K.; Foster, I. Cost-Aware Cloud Profiling, Prediction, and Provisioning as a Service. IEEE Cloud Comput. 2017, 4, 48-59.

[10] https://www.accenture.com/_acnmedia/PDF50/Accenture-Edge-Analytics-Mobility.pdf

[11] Matthews K, 4 challenges real time data still faces in 2018. https://www.rtinsights.com/4-challengesreal-time-data-still-faces-2018/, 2018