The past few decades, cloud computing has been used because of its high cost efficiency and flexibility which is achieved through consolidation, in which computing, network management, and storage functions work in centralized manner. Along with developmentof cloud computing, mobile Internet and Internet of things applications has also developed rapidly resulting in the current centralized cloud computing architecture to encounter severe challenges.Mobile devices try to obtain sophisticated applications from the distant centralized cloud which would result in high latency because of imposition of high loads on both radio access networks and back haul networks.Additionally, there has been an explosive growth in the number of end user demands and also there has been a great deal ofvariance in access devices.
New set of challenges has been introduced by emerging IoT, which cannot be addressed adequately by centralized cloud computing architecture such as, uninterrupted services with intermittent connectivity, capacity constraints, stringentlatency, resource-constrained devices and enhanced security. Edge computing, an advanced cloud computing paradigm breaks through these challenges is required12.There are certain key characteristics of edge computing on why it is beneficial in IoT.Faster response time between device and application server At the application level, the metric of interest would be the round trip time (RTT app ).Applications can serve many mobile endpoints simultaneously. Fast coordination and control is required in mobility which would not be achievable with a distant data center. These problems can be solved by computing resources at the edge.
Example of fast coordination and control requirements would be for fleet of drones and robots in precision farming. Another example would be autonomous and assisted driving cars with lane changing maneuvers, around the corner vision and hazard warning. In smart grid control applications and real time oroperational analytics of sensor data would require ultra-low latency request/ response loops, which could be provided by edge computing.
Applications where edge computing would be beneficial would be cloud gaming where metrics like fairness would be optimized with dynamic placement of game servers. Certain application would require the RTT app to match the expectations of the human brain for example in tactile and virtual reality (VR) applications. The RTT app for Oculus VR users should be under 20ms, if these expectations are not met it would result in motion sickness.
Smart glasses, computer vision with deep neural networks and synthesis of step by step instructions for cognitive-assistance applications would require round trip time to be under 600ms.Better battery lifetime, reduced device weight and improved form factorAlthough there has been improvement in computing capabilities of mobile devices, wearable watches and other IoT devices , they are still fundamental challenges which constrain their growth such as memory size,heat dissipation, battery life, etc.12 Mobile devices can offload heavy computation to the node at the edge to reduce consumption of power and allow them to be lighter and smaller and eventually improve battery life and form factor because of offloading energy consuming computations to the edge .This would apply for drones, wearables like smart helmets and head mounted devices(HMDs).More data storage and computation power for constrained devices Along with offloading the heavy computations, devices can also offload data to nearby edge node with more powerful resources such as GPU, Field-programmable gate arrays (FPGAs) and memory 13.
Bandwidth savingsFor an enterprise sending all data generated from sensors and devices to a distant private or a public cloud would prove to be expensive and wasteful, and the enterprise has to invest heavily on transportation infrastructure. Instead the huge data can be processed at the edge. Examples would include analytics for sensor data and location based services for 3D indoor navigation. Faster response time and bandwidth savings would be critical in automotive navigation application for generating and dispersing lane level maps and in event venues like stadiums for orchestrating the videos from multiple cameras possibly combined with advertising.Enhanced privacyAs IoT technology becomes the industry norm in markets from banking to health-care,it is vital that the data captured by the IoT devices to be protected as well. In certain scenarios, the data fusion, real time analytics and the generated actionable knowledge must be performed close to the source of the sensitive data to ensure privacy. For example in hospitals where high frequency data generated from intensive care unit machines can be fused with low frequency data from cloud based medical health record system to generate meaningful insights that may improve patient care 13. Another example would be in enterprise video surveillance networks, where face database of the employees would be fused with video captured data with images for granting access to a particular level.
Better cybersecurity and improved service availabilityAt the network edge, various security policies can be enforced for example to prevent the spreading of distributed denial of service attacks and data center protection. The traditional cloud services are although highly available, making use of edge nodes would help to achieve independence and autonomy from connectivity to main clouds 13.