Abstract— For the past few years Internet ofThings has been revolutionising our daily lives by connecting everything aroundus. It is the network of physicaldevices, vehicles, home appliances, and other items embedded with electronics,software, sensors, actuators, and network connectivity which enable theseobjects to connect and exchange data. Currentlymost interaction between the IoT devices and the back-end servers is donethrough large scale cloud data centres.
However, with the exponential growth ofIoT devices and the amount of data they produce, communication between “things”and cloud will be costly, inef?cient, and in some cases infeasible. In thispaper, we have proposed a new computing paradigm, named Edge Mesh, whichdistributes the decision-making tasks among Edge devices within the networkinstead of sending all the data to a centralized server. All the computationtasks and data are shared using a mesh network of Edge devices and routers.
Edge Mesh provides many bene?ts including distributed processing, low latency,fault tolerance, better scalability, better security and privacy, etc. Thesebene?ts are useful for critical applications which require higher reliability,real-time processing, mobility support, and context awareness. We ?rst give anoverview of existing computing paradigms to establish the motivation behind EdgeMesh. Then, we describe in detail about the Edge Mesh computing paradigmincluding the proposed software framework, research challenges, and bene?ts ofEdge Mesh. Keywords: Fog Computing, Mesh Network, Cloud Computation.I.
IntroductionThe Internet of Things (IoT) is likelyto be incorporated into our daily life, in areas such as transportation,healthcare, industrial automation, smart home, and emergency response. The IoTenables things to see and sense the environment, to make coordinated decisions,and to perform tasks based on these observations 1. In order to realize thefull bene?ts of the IoT, it will be necessary to provide suf?cient networkingand computing infrastructure to support low latency and fast response times forIoT applications. Cloud Computing has been seen as the main enabler for IoT applicationswith its ample storage and processing capacity.
Nonetheless, being far fromend-users, cloud-supported IoT systems face several challenges including highresponse time, heavy load on cloud servers and lack of global mobility.In the era of upcoming high datarequirement applications, it may be inef?cient to send the extraordinarilylarge amount of to the cloud, due to the high cost of communication bandwidth,and due to the high redundancy of data. Instead of moving data to the cloud, itmay be more ef?cient to move the applications and processing capabilitiescloser to the data produced by the IoT.
This concept is referred to as fogcomputing.Edge Mesh proposes the idea of usingEdge devices to enable distributed intelligence in IoT. In 2, distributedintelligence is de?ned as “cooperation between devices, intermediatecommunication infrastructures (local networks, access networks, globalnetworks) and/or cloud systems in order to optimally support IoT communicationand IoT applications”. They consider that distributed intelligence involvesboth processing and networking elements. We consider distributed intelligencein a broader perspective which involves everything from data analytics andnetworking to other functionalities such as data management, device management,resource management, service management, orchestration, etc.
There are manyresearch questions to be answered to develop Edge Mesh such as: How to de?nethe network and computing model? How to distribute processing of data? How tojointly optimize communication and computation? Etc.II. CLOUD COMPUTINGCloud computing can be defined as the practice of using anetwork of remote servers hosted on the Internet to store, manage, and processdata, rather than a local server or a personal computer.
IoT devices are resource constraint whichlimits their applicability; on the other hand, Cloud has abundant resources.Therefore, IoT can make use of resources in the cloud to make up for itslimited resources 4. Cloud can bene?t IoT in many ways includingcommunication, computation, and storage. Data collected by IoT devices can bestored in cloud and processed in a low-cost and effective manner. Cloud isespecially useful for IoT applications that are computation intensive and/oruse data driven processing 4. Cloud computing paradigm is heavily dependenton Internet connectivity.
Due to intermittent network connectivity, networklatency becomes high which is not suitable for applications with the real-timerequirement. As the amount of data being generated by IoT devices is becominghuge, it is very dif?cult to send all the data to Cloud due to limitedbandwidth constraint. There is also a major issue of security and privacy asdata travels along intermediate networks which can be prone to attacks and ifthe data is stored at public cloud then chances of unwanted access and/orcompromise becomes higher.
Since Cloud data centers are usually located at afaraway place, latency is higher which means Cloud computing paradigm is noteffective for an application that requires mobility support.III. FOG COMPUTINGCloud computing paradigm suffers fromfour major issues discussed above, i.e.
latency, security, privacy, andmobility, which has motivated researchers to propose a Fog Computing paradigm.Open Fog Consortium de?nes Fog Computing as “a system level horizontal architecturethat distributes resources and services of computing, storage, control andnetworking anywhere along the continuum from cloud to Thing” 5. Fog Computingis a distributed paradigm that provides characteristics such as low latency,real-time interaction, distributed analytics, context awareness, geographicaldistribution, mobility support, which are not supported by centralized Cloudcomputing paradigm 6. Fog Computing shares many similarities with EdgeComputing paradigm as both allow computation closer to devices that produce data78.
Fogcomputing focuses on infrastructure perspective while Edge computing focuses onthings perspective 8. There is another similar paradigm called Mobile EdgeComputing, which was proposed by ETSI as a platform that pushes cloud computingcapabilities closer to mobile devices in radio access networks (RAN) 9. Thereare many challenging issues that need to be resolved to realize the fullpotential of fog computing paradigm. These issues are related to fognetworking, Quality of service, interfacing and programming, computationof?oading, accounting, billing, monitoring, provisioning and resource management,and security and privacy 9.
IoT integrates different application domains whichcan have varying requirements. Fog computing and cloud computing paradigms bothprovide different bene?ts but they can work complementary with each other tosatisfy multiple applications requirements 6. Currently researchers are workingto integrate Fog and Cloud computing paradigms 10 11. IV.EDGE MESHIn this section, we give details about theEdge Mesh computing paradigm, which aims to enable cooperation betweendifferent types of devices and enable distributed intelligence in IoT. EdgeMesh can be de?ned as a computing paradigm that uses mesh network of Edgedevices and routers to enable distributed decision-making within the network. Components ofEdge Mesh:1) End Devices: End devices arethose devices which have the capacity to sense the surrounding and change itbased on the requirement.
End devices are responsible for sensing andactuation. Devices in a smart home such as camera, lights, thermostat, etc. aresome examples of End devices.2) Edge devices: It is any computing or networkingresource residing between data sources and Cloud-based data center.
In ourcase, we consider Edge devices as those devices which are either connected toend devices or to Cloud. These devices are responsible for decision-making andenabling interaction between End devices. Any device that can be used forprocessing and enables connection between different end devices can be used asEdge device. 3) Routers: Routers are used for relaying data between edge devices.Their function is just to route the data. Routers are not used for processingor enabling decision-making like Edge Devices. Routers and Edge Devicestogether form a mesh network which is used for sharing computation and dataamong Edge devices. 4) Cloud: Cloud provides abundant computing resourcesincluding networks, storage, processing, application, services, etc.
Traditional IoT systems use a centralized Cloud server for enablingdecision-making. However, in the case of Edge Mesh, major decision-making isdone by Edge devices instead of Cloud. Cloud is integrated with other devicesonly to be utilized for very speci?c application requirements that cannot bemet using Edge devices. TABLE IComparison between traditional and edge mesh network Traditional Mesh Network Edge Mesh Distribute data from one node to another within the mesh network. Used for distributing data within Edge Mesh as well as enabling interaction between other devices including end devices and Cloud Nodes in traditional mesh network just make routing decisions. Besides routing, Edge devices in Edge Mesh make decisions for different computation tasks including processing, storage, networking etc. Nodes in traditional mesh do not make decisions regarding end devices Edge devices make decisions regarding interaction between end devices.
Nodes in traditional mesh are not responsible for managing data shared between end devices Edge devices in Edge Mesh are responsible for managing data shared between end devices IV. BENEFITS OF EDGE MESHBene?ts of Edge Mesh The mainobjective of Edge Mesh is to enable distributed intelligence which helps EdgeMesh to provide bene?ts associated with distributed computing systems. Suchbene?ts include fault tolerance, better scalability, and ef?cient performancedue to the distribution of load. There are some other bene?ts provided by EdgeMesh by the virtue that it integrates characteristics from three differentcomputing paradigms, i.e. Cloud Computing Fog Computing, and CooperativeComputing.
Edge Mesh provides the best features of the three computingparadigms. The bene?ts provided by such integration include low latency, betterservices, and higher security and privacy.1) FaultTolerance: Edge Mesh provides fault tolerance in terms of both communicationand computation. Since a mesh network is used for distributing data amongdifferent devices, it provides many redundant connections.
In the case offailure of a device in the communication path, other paths can be used fordistributing data. Edge Mesh also provides redundancy for computation tasks.The responsibility of any computation task lies on multiple Edge Devices thatcooperate with each other, therefore, failure of a single device does notjeopardize the whole system. 2) Scalability:Scalability is an important requirement for IoT systems as the number ofdevices will continue to increase in the coming future.
A computing paradigm thatrelies on the centralized server for computation tasks cannot be scaled up.Edge Mesh, on the other hand, has been proposed to enable distributedintelligence which makes it suitable for IoT applications. Edge Mesh isdistributed so all the data is not sent to a single Edge device.
The data issent to multiple Edge devices which can then share data so the communicationbottleneck issue is resolved due to the distributed nature of the system.3) LoadDistribution: Computation tasks can be of?oaded to other Edge devices whichspeed up the processing time. A single Edge device is not overloaded whichusually leads to better performance.
Edge Mesh distributes the load among Edgedevices which leads to better response time, reduced makespan, and higherthroughput. Distribution of load also makes the systems more ?exible, i.e. inthe case of device failure, other devices can share the load of failed device.IoT systems are dynamic, as devices can be mobile, added, removed, or changedin con?guration. Edge Mesh can adjust to such changes as Edge devices cancooperate with each other.4) LowLatency: Many IoT applications such as healthcare, video analytics, autonomousvehicles, traf?c management systems, emergency response systems, smart parking,etc. have low latency requirement.
Cloud computing paradigm is not ef?cientenough to be used for these time-critical applications. A large portion of thetime is consumed to transfer the data to and from a remote server which doesall processing tasks. Edge Mesh uses local Edge devices which can performcomputation tasks and share data within the required deadline. V. APPLICATIONS The application scenariosdiscussed in this section are related to three different application domains,Smart Home, Intelligent Transportation System, and Healthcare. Theseapplication scenarios help in illustrating the bene?ts of Edge Mesh and give anunderstanding of scenarios where Edge Mesh computing paradigm can be ofsigni?cant use.
Smart Home has been one of the oldestapplication domains of Internet of Things. The main objective of Smart Home isto improve the comfort level, security, and safety of people inside the homewhile considering energy conservation and cost into account. Intelligent transportationsystem (ITS), also referred to as Internet of Vehicles (IoV), is another importantapplication domain of IoT. The technological advancement in sensor technologiesand vehicular communication technologies which is a communication protocol toenable data exchange between high-speed vehicles and between vehicles androadside infrastructure units, has enabled exciting applications for ITSdomain. Google and other big corporations are now working on autonomous andconnected vehicles to improve road safety, traf?c ef?ciency, and enable otherservices such as intelligent parking, accident prevention, collision warning,etc. VI.
OPEN CHALLENGESThe two main characteristics of EdgeMesh are that it uses distributed computation and integrates characteristicsfrom different computing paradigms. However, there are various open challengesin it, Edge Mesh should support communication between different types ofdevices such as Edge device, End device, routers, and Cloud. The data should notonly be shared between different types of devices but must be understood by thedevices. Communication protocols used in IoT suffer from low data rate,frequent packet losses which make it dif?cult to achieve reliablecommunication. Edge Mesh enables distributed intelligence;however, where the intelligence should be placed is a major research question.There are many other research questions too including, How the devicescooperate with each other? Which devices should share data? How do devicesdecide which data to be shared? Who decides the distribution of tasks amongdevices? Which factors determine the distribution of tasks? etc.
Thecomputation tasks are usually distributed among Edge devices, but Cloud canalso be used for big data analytics on large historical data. So, the tasks canbe distributed among different devices depending on application requirement andresources available on the devices. It is challenging to determine how thetasks must be distributed among Edge devices as it requires joint optimizationof computation and communication. Edge devices need to take care of many issuesincluding access control, resource allocation, QoS, security, data conversion,data management, etc. This requires Edge devices to be robust and ?exible. Itis challenging to manage all the tasks simultaneously using Edge devices asthese devices are heterogeneous, resource constraint, and distributed. Thechallenges in implementing Edge Mesh are a combination of many factorsincluding wireless distributed computing issues, IoT related challenges,embedded device constraints, software implementation issues, theoreticalmodelling limitations, algorithmic challenges, issues related to distributeddata analytics, etc.
VI. CONCLUSIONThis paper proposes a new computingparadigm, Edge Mesh, which focuses on enabling distributed intelligence in IoT.Edge Mesh distributes the whole application into sub-tasks which aredistributed among Edge devices. Edge devices together with routers form a meshnetwork which is responsible for many computation tasks such as storage,processing, data sharing, etc. Edge Mesh tries to integrate best features fromCloud computing, Fog computing, and cooperative computing to providemulti-dimensional features.
Many research challenges discussed in this papersuch as enabling meaningful exchange of data between heterogeneous devices, proposingdistributed security and privacy algorithms for resource constraint devices, enablingintegration of different computing paradigms, etc. are all open issues thatrequire more research efforts. References1 A. Al-Fuqaha, M. Guizani, M.Mohammadi, M.
Aledhari, and M. Ayyash, “Internet of things: A survey onenabling technologies, protocols, and applications,” IEEE CommunicationsSurveys Tutorials, vol. 17, no. 4, pp. 2347–2376, 2015.2 C.
-W. Tsai, C.-F. Lai, M.-C.Chiang, L. T. Yang et al.
, “Data mining for internet of things: A survey.” IEEECommunications Surveys and Tutorials, vol. 16, no. 1, pp.
77–97, 2014.. 3 F. Bonomi, R. Milito, J. Zhu,and S. Addepalli, “Fog computing and its role in the internet of things,” inProceedings of the First Edition of the MCC Workshop on Mobile Cloud Computing,pp. 13–16, 2012M.
Wegmuller, J. P. von der Weid, P. Oberson, and N. Gisin,”High resolution fiber distributed measurements with coherent OFDR,” in Proc.
ECOC’00, 2000, paper 11.3.4, p. 109.4 A.Botta, W.
De Donato, V. Persico, and A. Pescap´e, “Integration of cloudcomputing and internet of things: a survey,” Future Generation ComputerSystems, vol. 56, pp.
684–700, 2016. 5 OpenFog, “De?nition of fogcomputing,” http://www. openfogconsortium.org/resources/#de?nition-of-fog-computing,accessed: 2017-05-14. 6 F.Bonomi,R.Milito,J.Zhu,andS.
Addepalli,”Fogcomputingandits role in the internet of things,” in Proceedings of the ?rst edition of theMCC workshop on Mobile cloud computing. ACM, 2012, pp. 13–16. 7 F. Bonomi, R. Milito, P. Natarajan, and J.
Zhu, “Fog computing: A platform for internet of things and analytics,” in BigData and Internet of Things: A Roadmap for Smart Environments. Springer, 2014,pp. 169–186. 8 W.Shi,J.Cao,Q.Zhang,Y.
Li,andL.Xu,”Edgecomputing:Visionand challenges,” IEEE Internet of Things Journal, vol. 3, no. 5, pp. 637–646,2016.
9 S. Yi, C. Li, and Q. Li, “A surveyof fog computing: concepts, applications and issues,” in Proceedings of the2015 Workshop on Mobile Big Data. ACM, 2015, pp. 37–42.10 M. Villari, M.
Fazio, S.Dustdar, O. Rana, and R.
Ranjan, “Osmotic computing: A new paradigm foredge/cloud integration,” IEEE Cloud Computing, vol. 3, no. 6, pp. 76–83, 2016. 11 A.Munir,P.Kansakar,andS.
U.Khan,”Ifciot:Integratedfogcloudiot architectural paradigm for future internet of things,” arXiv preprintarXiv:1701.08474, 2017. 22 T. H.
Luan, L. Gao, Z. Li, Y.
X12 F. Bonomi, R. Milito, P. Natarajan, and J.Zhu, “Fog computing: A platform for internet of things and analytics,” in BigData and Internet of Things: A Roadmap for Smart Environments.
Springer, 2014,pp. 169–186.