Edge Computing and the Future of Cloud
Future of Cloud and Edge Computing Within the Internet-of-Things
The internet of things (IoT) is evolving rapidly. In the future, it will connect billions or even trillions of devices. Most of these devices will be the ones that have not been considered for automation previously. This will open a huge market for developing new applications. The edge computing IoT devices will not be limited to previously launched devices such as Apple watches, Oculus Rift helmets, Google Nest, Fitbit Sports trackers, or Google Glasses.
Edge IoT devices can be of any kind that have been fitted with a chip and a sensor. It will not matter who the manufacturer is, as long as they have a chipset. Many applications can be developed to connect these devices to IoT and enable smart living, smart businesses, smart cities, etc. Currently, the data from these edge devices is hosted where its application server is located. This is usually one of the small number of Application Service Providers (ASPs) including Google, Amazon, Microsoft, Facebook and Apple.
The current centralized cloud computing architecture favors big data centers. It has been very successful so far. Customers can get data storage customized to their needs at a very small fraction of what it would cost them to host the data in-house. It also makes it easy to train AI by making available big data in one location.
But, this current model may prove to be insufficient to cater to IoT. The large volume and speed of data generation from IoT devices may make it difficult to transfer it to remote data centers in the required timeframe. This may lead to applications involving end-to-end communication experiencing high latency.
Local Data Storage
With an open edge cloud infrastructure, this challenge can be met by providing a local computing, storage and networking resource for IoT edge applications. These applications can store their data locally, where it can be processed. And, only this processed low-volume data can be sent to the remote data center. Moreover, if an IoT device application gets overloaded, it can transfer some tasks to the local cloud storage, decreasing the latency to a management delay. An open cloud strategy will also keep the room for innovation open.
Edge cloud computing is drawing increasing attention from the IT industry. However, the technology is still maturing and there are many challenges to be overcome. Most current edge computing frameworks involve physical computing servers that communicate with edge sensors for computation and storage or dockers that provide limited visualization support at the edge. These are standalone deployments mostly that perform tasks such as video surveillance or video analytics. Network Function Virtualization (NFV) and Software Defined Networking (SDN) for the future edge cloud are still in their early stages.
State of the Art
Here are some interesting projects that are being worked on by the academia and the industry:
A research group in Carnegie Mellon University is working on a project named Cloudlet. The goal of this project is to achieve the convergence between Mobile computing and cloud computing. They hope to achieve this through a multi-tier hierarchical structure.
A Fog Computing concept was proposed by Cisco in 2012. The project aims to extend cloud computing and services to the edge of the network to ease the wireless data transfer, and provide data, compute, storage, and services to the end-users with proximity, dense geographical distribution, and mobile support.
The Industry Specification Group (ISG) from the European Telecommunications Standard Institute (ETSI) launched a Mobile-Edge Computing (MEC) Initiative in 2014. Its goal is to provide a network architecture that enables cloud computing capabilities and an IT service environment at the edge of mobile cellular networks for the future 5G and IoT world;
Central Office Re-architecture as a Datacenter (CORD)
AT & T and ON.lab launched a joint project named Central Office Re-architecture as a Datacenter (CORD) in 2017, that is focused on transforming legacy Central Offices (C.O.) into CORD which integrates NFV, SDDN, and Cloud into services providers’ access networks;
Nebula, along with some other projects, is being funded by the Future Internet Architecture (FIA) program of the National Science Foundations (NSF). It features a distributed edge cloud infrastructure that allows volunteer nodes from the edge to contribute to MapReduce tasks for data-intensive computing;
FemtoCloud is a project by a research group from Georgia Institute of Technology. It groups mobile devices so they can be controlled as a cluster. The idle computing resources from these mobile devices can be shared within this group for a specific task managed by a controller;
Mobile-edge Offloading and Foraging
Mobile-edge offloading and foraging technologies for the collaboration and coordination between the mobile devices and the edge servers are being developed by Duke University (MAUI Project), Intel Labs Berkeley (CloneCloud), University of Southern California (Odessa), and University of Michigan (COMET). The aim of these projects is to use mobile-edge offloading to allow the mobile devices to offload computation and storage of the edge servers for those tasks that need more resources.
Technologies Supporting Edge Cloud and Computing
Here are the four key enabling technologies for IoT applications:
NFV technology is being used to enable a small-sale cloud computing platform in order to push computing, storage and networking resources to the edge and enable future IoT applications;
SDN technology is being used to network, configure, control and manage NFV in the edge cloud. This can lead to cost reduce. It can also increase flexibility and programmability of the VNFs in the edge cloud.
Automated Orchestration is being used to select, deploy, monitor and control the configuration of hardware and software resources for application delivery.
Dynamic Offloading to enable the orchestrator to effectively work with various types of IoT devices to offload their data and computation to the edge cloud and dynamically and optimally commit appropriate resources to carry out these offloaded tasks matching the demands.
IoT Applications Benefiting From Edge Cloud and Computing
Here are three types of future IoT applications that cloud benefit directly from the edge cloud:
Applications That Require Low Latency
The centralized data architecture will not be enough for such applications due to two reasons. Firstly, because of the large data volume being generated. Secondly, because of the long distance between the origin of data and the host data center. This includes video recording and gaming applications that require latency as low as 10 milliseconds. This becomes all the more important in industries such as healthcare and transportation where a delay in response can be life threatening.
Applications Requiring High Data Bandwidth
Applications, such as those supporting Augmented Reality (AR) and Virtual Reality (VR), generate a lot of data that require a high bandwidth. As data getting produced continues to increase, it will become necessary to first process the data through the edge cloud and computing to reduce its volume before sending it to the remote data center.
Applications Supporting a lot of IoT Devices With Limited Capabilities
IoT devices that are situated at the periphery support only a limited computing power and battery life. The limitation of hardware makes it hard to perform power intensive computations for these devices. It is far better to offload these computations to the edge cloud.
Challenges Facing Edge Computing
Here are some key issues facing the development of Edge Cloud and Computing state of the art:
- NFV and SDN Synchronicity: Even though, NFV and SDN technologies can really benefit edge cloud commuting, the technologies are not mature yet, and more research and development is required to synchronize them seamlessly;
- 4G/5G Mobile Networks or Internet: Depending on the network architecture the implementation and location of edge cloud can vary significantly. This will also mean different corresponding participating vendors;
- IoT Devices and Edge Coordination: For edge servers to facilitate IoT devices with computing, storage and networking, devices and the edge must coordinate with each other;
- Southbound Interfaces (SBIs) and Northbound Interfaces (NBIs): SBIs and NBIs are two important types of interfaces to make everything happen seamlessly. However, multiple stakeholders involved in NFV and SDN integration may not be motivated to work together and provide clear interfaces. This makes it difficult for SBIs and NBIs to function;
- SDN Multi-cloud Scenarios: SDN is going to be replaced by Software-Defined Exchange (SDX). This will enable inter-domain networking. Moreover, large-scale interconnections of Software-Defined Internet (SDI) owned and operated by various organizations will also be enabled. The SDX technology may penetrate edge cloud as well.
- Security: Security is the most important challenge for edge cloud and computing. Involvement of multiple technology multiples security risk as well. Since the future edge clouds will enable more programmability and the open platform will allow more third-party software and hardware vendors to weigh in and contribute, it is important to control and manage the potential risks among different stakeholders. Also, appropriate authentication, authorization, and auditing mechanisms may be required to identify and protect the trusted parties. Moreover, they are required to defend from potential malicious attacks and misuses.
Edge Cloud and Computing is in its early stages, and faces many challenges. Supporting technologies such as NFV and SDN are promising to address some of these challenges. The effort currently being invested in the development of Edge Cloud is definitely worth it because it will enable us to truly enter the era of IoT. Of course, the applications of IoT for businesses across industries are vast, and are only limited by imagination. As always, Croyten is staying one step ahead of the technology curve. Our researchers are already looking for ways to include IoT application development in our list of services.