您的瀏覽器不支援JavaScript語法,但是並不影響您獲取本網站的內容。

5G通訊網路之邊界與霧計算整合

107-04-30

ITRI ICL Lab / Samer Talat, Shahzoob Bilal Chundrigar

In 5G era, developing an innovative architectural network paradigm is essential to address the technical challenging of 5G application’s requirements in a unified platform. In particular, forthcoming applications will provide a wide range of networking, computing and storage capabilities closer to the end-users. In this context, the 5G-PPP Phase two project named “5G-CORAL: A 5G Convergent Virtualised Radio Access Network Living at the Edge” aims at identifying and experimentally validating which are the key technology innovations allowing for the development of a convergent 5G multi-RAT access based on a virtualized Edge and Fog architecture is scalable, flexible and interoperable with other domains including transport, core network and distant clouds. Through the on-going 5G-CORAL project, shopping malls, high-speed trains and connected cars testbeds in Taiwan and Italy will be used to validate the system in end-to-end large-scale, supporting innovative applications such as augmented reality, car safety, and IoT gateway.


Introduction

As the 5G applications and their technical requirements emerge, it is clear that certain applications such as augmented reality, connected vehicles, remote operations, and robotics will require very low end-to-end latency (~0.1-20 milliseconds) ‎[1]‎[2]‎[3]. This is extremely challenging and stressing for the network to deliver through a pure centralised architecture. The need to provide networking, computing, and storage capabilities closer to the end users has therefore arisen, leading to what is known today as the concept of intelligent edge.

On one hand, the European Telecommunications Standards Institute (ETSI) has been the first to address this need by recently developing the framework of Mobile Edge Computing (MEC) ‎[4]. Such an intelligent edge could not be envisaged without virtualisation, where ETSI has also been pioneering the concept of Network Functions Virtualisation (NFV) ‎[5]. In the ETSI MEC framework, the edge virtualisation substrate has been largely assumed to be fixed (stationary) and centralised in a pre-defined location (e.g., edge data centre), and exclusively owned by one stakeholder, that is the operator. A relaxation of these assumptions is inevitable to address challenging but practical scenarios where, for example, computing, storage and network connectivity can be provided by any node or device even on the move (e.g., onboard a car or a train) and no matter who its owners are. Needless to say, MEC provides IT and cloud-computing capabilities within the Radio Access Network (RAN) in close proximity to mobile subscribers which can reduce latency, ensure highly efficient networks operation and service delivery and offer an improved user experience. Moreover, the mobile device can offload computing tasks to MEC servers and fetch contents from MEC servers via RAN instead of doing such jobs to/from cloud servers via RAN and Core networks.

On the other hand, Fog computing is a system-level horizontal architecture that distributes resources and services of computing, storage, control and networking anywhere along the continuum from Cloud to Things. OpenFog consortium ‎[6] was recently established with the aim to leverage virtualisation for a cloud-to-thing continuum where applications can be distributed anywhere between cloud and things. In particular, Fog computing can provide an alternative to cloud computing that puts transactions and resources at the edge of the network, rather than establishing channels for cloud storage and utilization. Also, Fog computing reduces the need for bandwidth by not sending every bit of information over cloud channels and instead aggregating it at certain access points, such as routers. In addition, Fog computing facilitates the operation of computing, storage and networking services between end devices and cloud computing data centers.

Notably, integrate edge and fog will arise opportunities for networking functions to execute closer to the end users devices and things benefiting from inherent low latencies. Being in close proximity to the access, the edge and fog become thus an attractive natural place for hosting RAN functions, which could also be envisaged to be virtualised. Virtualised RAN functions may execute dynamically anywhere in the edge and fog, but also in distant clouds. This makes it possible to consider breaking the so far exclusivity of virtualised RAN functions to access nodes or base stations and envisage them for end-user devices and things too, so that some functions of these end devices may be moved into the edge and fog. End devices could also be envisioned to offer their own networking and computing capabilities to nearby devices and things, thus becoming a living component of the fog in what is a new paradigm of device-to-device (D2D) communications. Virtual Network Functions (VNFs) from the transport and core could also be hosted in the edge and fog so as to save bandwidth in their respective domains and offer local breakout where required. A rich set of context information from the RAN, as well as from other network domains, can be extracted. This context information could also be offered as services through the edge and fog for applications and virtualised functions. Furthermore, the consumed services need to be optimised to maintain 5G KPIs. Moreover, various Radio Access Technologies (RATs), such as 5G New Radio, 4G, WiFi and IoT (e.g., cellular NB-IoT, non-cellular BLE and ZigBee), are envisioned to offer connectivity to various types of devices and services in a given local access area. These RATs are therefore set to share and take full advantage of the edge and fog to better enhance their performance and cost-effectiveness. Virtualised RAN functions from these RATs could, therefore, find tailored hosting in the edge and fog. Context information collected from all the various access nodes and devices could also find in the Edge and Fog a perfect brokerage platform for offering their services to internal and external applications. An opportunity arises for tight coordination and cooperation between these RATs co-serving the same local access area. Such tight coordination comes with a stringent synchronisation requirement which could only be offered in an environment of very low latency. Unlike conventional approaches for multi-RAT access convergence, where the focus is mostly put on the harmonisation and integration of communication protocol stacks from different RATs, the 5G-CORAL project takes a unique approach by bringing together all these RATs into a common virtualised framework built on integrated edge and fog networking and computing resources. In 5G-CORAL, several challenges arise for our envisioned solution to address, such as:

Challenge 1

To develop a system model that includes use cases, requirements, architecture, and business models to design and validate the solution

R&D Topics

  • Identify and prioritise use cases, deployment scenarios, and requirements for the design and demonstration of 5G-CORAL solution.
  • Develop the 5G-CORAL architecture leveraging on existing industrial frameworks for NFV, SDN, MEC, and fog computing.
  • Develop the 5G-CORAL system framework for supporting convergence between the multiple RATs envisioned.
  • Define step-based procedures and techniques for enabling incremental deployment of 5G-CORAL solution into existing networks.
  • Develop business models involving all stakeholders of the 5G-CORAL value chain, such as operators, vendors, service/application/cloud providers, facility owners, end users, etc.

Challenge 2

To design virtualised RAN functions, services, and applications for hosting in Edge and Fog computing System (EFS)

R&D Topics

  • Explore the virtualisation of RAN functions in the EFS for multiple RATs, develop their requirements, and assess their merits from an access convergence viewpoint.
  • Specify EFS services for collection, aggregation, publishing, and use of radio and network context information by applications and possibly virtualised functions.
  • Develop EFS functions using EFS services from multiple RATs and the transport and core networks in support of access convergence
  • Develop EFS applications using EFS services from multiple RATs and the transport and core networks to improve network KPIs and user QoE, such as: IoT gateway, augmented reality, user-targeted advertisements, and cars communication.

Challenge 3

To design an Orchestration and Control System (OCS) for dynamic federation and optimised the allocation of EFS resources

R&D Topics

  • Extend existing industrial frameworks for NFV, MEC, and fog to best suit dynamic environments where EFS resources are volatile.
  • Develop federation mechanisms for EFS resources belonging to multiple owners and subject to different technical, business, and administrative requirements.
  • Develop interfaces for automated deployment of EFS functions and applications.
  • Integrate the EFS with central clouds to enable instantiation and migration of virtual functions and applications between the EFS and central clouds.
  • Develop orchestration and control algorithms for elastic placement and migration of EFS functions and optimised allocation of EFS resources.

Challenge 4

To integrate and demonstrate technologies in large-scale testbeds making use of facilities offered by Taiwan, and measure their KPIs

R&D Topics

  • Customise existing testbeds in Taiwan to meet the needs of 5G-CORAL proof-of-concept in large-scale deployments.
  • Integrate and validate EFS and OCS in large-scale testbeds, such as shopping mall, high-speed train, and connected cars.
  • Demonstrate and trial multi-RAT access convergence and low latency applications, such as augmented reality and car safety, in real-world scenarios involving real users.
  • Evaluate the performance of 5G-CORAL solution in the field through measurement of relevant KPIs on data rates and latency in low, medium and high mobility environments.

5G-CORAL Architecture

The design of the integrated Fog, Edge and Cloud system, namely 5G-Coral architecture, follows the ETSI MEC and ETSI NFV concepts and envisages a mix of physical and virtualised resources available on the Fog and Edge devices to form an ETSI NFV compliant infrastructure. ‎[7] presents the initial idea of the proposed architecture.

The proposed solution contemplates two major building blocks, namely (i) the Edge and Fog computing System (EFS) subsuming all the edge and fog computing substrate offered as a shared hosting environment for virtualized functions, services, and applications; and (ii) the Orchestration and Control System (OCS) responsible for managing and controlling the EFS, including its interworking with other (non-EFS) domains (e.g., transport and core networks, distant clouds, etc.).

Figure 1: 5G-CORAL Architecture

Figure 1: 5G-CORAL Architecture

The EFS building block is formed by three main elements: i) EFS Applications, ii) EFS Functions and iii) EFS Service Platform. EFS Applications perform computing and/or networking tasks for the end user or a third party. EFS Applications can consume and/or publish information through one or more services. Example of EFS Application might be an augmented reality application or a Fog Node Coordinator. EFS Functions perform computing and/or networking tasks required for network operations and/or performance enhancements. An EFS Function can consume and/or publish information through one or more services and can interact with one or more applications. Examples of function are a base band unit (BBU) and a mobility entity management (MME). Finally, EFS Service Platform provides information through a publish/subscribe or request/reply system. This information can be published or consumed by functions and/or applications which are either part or external of the EFS. Examples of EFS Service is the radio network information service from ETSI MEC. Such service could be provided either by a virtualized Base Band Unit (vBBU) through the E2 interface or by a physical eNodeB through the T8 interface.

According to the ETSI NFV architecture, each of the EFS Applications, Functions and Services may have an Element Manager which is in charge of applying the configuration and management policies as defined by the EFS Manager at the OCS (interface O6, functionally similar to ETSI NFV Ve-Vnfm-em ‎[8]). The Element Manager of the EFS Service Platform partially plays the role of the ETSI MEC Platform Manager ‎[9]. Its scope is limited to service-related management only (e.g., policies configuration) and is not in charge of the lifecycle management of the applications and functions which is instead the responsibility of the OCS. The different EFS components are designed to run in an amalgam of devices, with different capacities and features. EFS functionality will partially reside in Fog devices distributed near the user, access nodes, incorporating multiple technologies and virtualized infrastructure in servers located at the Edge.

The OCS design follows the ETSI NFV architecture and comprises the following components: i) EFS Orchestrator, ii) EFS Manager and iii) Virtualised Infrastructure Manager (VIM). As in the ETSI NFV architecture, the Orchestrator takes care of building the End to End service, by requesting the underlying modules to instantiate the required VNFs, EFS Applications and Functions. The EFS Manager manages the lifecycle of the Applications, Functions, and Services, including migration and on-demand scaling. Finally, the VIM is in charge of managing the virtualized infrastructure as in ETSI NFV architecture. Operations in the scope of the VIMs are keeping an inventory of available resources (including their discovery) and managing the connectivity of the various Applications, Functions, and Services. The key functionality of OCS is to build a coherent view of the EFS capabilities in a certain area considering the different resources and access technologies deployed.

Use Cases and Demonstration Platforms

A.Use Cases Examples

There are many services that could benefit from being hosted in the near proximity of the user. In combination with visibility into distributed private fog networks to more centralized edge networks. Some of the leading use cases are discussed as follows:

Safety in Connected Cars: Cars and other vehicles may exploit 5G and legacy wireless communications to improve traffic safety, to support drivers with real-time information on road and traffic conditions, and to make the mobility of emergency vehicles such as ambulances and fire trucks safer. This information is received by the vehicles nearby and can be used to alert all the drivers about what is going on the road. Such category of communications includes different ways: a vehicle to vehicle (V2V) or vehicle to infrastructure (V2I) communication, a communication with vulnerable road users, e.g. pedestrians and cyclists (V2P). The related applications can be hosted either distributed on the edge and fog systems , to optimize the network performance such as offloading and end-to-end network latency, then taking full advantage of the low latency communications at the EFS.

Smart City Services: In smart cities, jointly orchestrated edge and fog computing system can enhance multiple use cases such as security cameras vision and face recognition offloading, and smart traffic light system. In particular, security cameras vision and face recognition specific use case using of edge and fog systems to offload face recognition and computer vision algorithms to a resource-sufficient cloud environment, taking advantage of low latency communication and high bandwidth environment that the edge/fog provides. Furthermore, offloading to the joint edge and fog system will increase the number of connected devices as the price per device will greatly reduce. In addition, there is the smart traffic light system use case, where street lights jointly with cameras control the traffic in a certain region, in order to maintain a continuous flow of pedestrians and vehicles. Additionally, it could be able to prevent accidents and collect traffic statistics in order to improve the system globally.

Wireless High-speed Train Service: In the high-speed train, wireless train service can be optimized by utilizing the edge units. In particular, end users can benefit from cached information on the onboard edge server where required data is collected based on user demands. When a train is stationed, the onboard edge server exchanges (adds/deletes/updates) information with the updated one from on-land Server located in every station. The fact of having computational and networking resources locally in the train and their capacity to update and cache information on the stations open the door to novel applications taking benefit of the locality of the data. Examples of these applications are smart tourism guides (city information, 3D interactive offline city maps, points of interest), updated news (also based on the location of the train) or interactive train services (customer service, emergency, re-routes based on user’s destinations).

Robot and Augmented Reality in Shopping Mall: Various tasks in a shopping mall could be handled by robots to enhance experiences of the customers. For examples, we may have (1) the robots that guide customers to the shops of interest, (2) the robots that help customers to fetch merchandises, (3) the robots that provide personal assistance to the disabled customers, (4) the robot security guards, and (5) the robots that maintains and performs operational tasks on behalf of the shop owners. Many of these potential use cases require tight cooperation among robots, or even between robots and humans. Conventionally, cloud robotic could be employed, in which multiple sensors distributed across the shopping mall building collect raw data for cloud processing, to derive the instructions for the robots. Nevertheless, in some cases extremely low latency is needed and cloud robotic may not be feasible due to intolerable delay. Also, augmented reality (AR) applications can be deployed in the shopping mall to improve the user experience of the customers and the workers. For example, (1) Live Navigation, to provide accurate and live navigation to the destined shop, (2) Repair and Maintenance, where AR can be used to overlay instructions to reduce error-rates, (3) Connecting remote workers, allows an expert or the shop owner to assist his workers to complete the task remotely, (4) Additionally shoppers can also use the AR to see how the object will look like.

Industrial IoT: The current industrial IoT solution is based on a client/server design where client SWs are running on the devices or gateways and server SWs are running in the cloud. The latency becomes an issue for many real-time industrial IoT applications. For example, wireless robotics can require ms-level of latency. Edge and fog computing provides the opportunities to provide a low latency computing infrastructure to end-users. The fog nodes can host the light-weight computing and networking tasks with the lowest possible latency. The edge nodes usually have more computing, storage, and networking resources and thereby can act as a nearby small-scale cloud providing heavier services like data analytics, and coordinate and management of a huge amount of IoT devices. It may also provide the users additional values when data from different IoT networks are available and thereby can be co-processed in the edge and fog. For example, data from the presence sensors may be useful to improve the accuracy of the localization service, or to dynamically turn on/off resources at the edge to save energy.

B.Demonstration Platforms

This subsection highlights the demonstration platforms, where appropriate use cases will be selected to demonstrate in the following listed real-world platforms.

Connected Cars: The testbed is located in Turin, Italy, and coordinated by Azcom and Telecom Italia. The connected cars testbed will support the trial for those use cases that exploit the connected car's communication through the cellular infrastructure (V2I). In this scenario, data regarding speed, direction and position, are collected from vehicles and can be used in a different way for improving the cars safety. In fact, the V2I communications add significant value to the advanced driver-assistance systems in terms of improved safety, traffic optimization and improved driving experience. Within the suggested testbed setup, it could be possible to consider a number of different end-user applications such as roadworks reporting, weather conditions reporting, emergency vehicles approaching, position tracking and collision avoidance.

Shopping Mall: The testbed is located in Taipei, Taiwan and is organized by ITRI. It provides facilities for performing experiments and pilot deployments in realistic dense scenarios, both regarding infrastructure and users. The testbed also allows users to take part in the demonstration trials. Computation offloading, network offloading and mission-critical services will be demonstrated in a multi-RAT dense environment. For instance, AR, Robotics will benefit from location and multi-RAT context information. Also, IoT gateways and AR application will take advantage of the vicinity of computing resources for offloading heavy processing tasks from end-user devices/sensors to the edge/fog computing devices to leverage the low latency communications.

High-speed Train: The testbed is located in Hsinchu, Taiwan and is coordinated by ITRI. It is amongst very few commercial high-speed train testbeds in the world capable of collecting and experimenting real high-speed data on real scenarios. The envisioned goal of this testbed is to verify seamless connection in the high-mobility scenario. In this high mobility case, the seamless connection can be achieved through breakout the traffic and deployment of mobility functions on the edge/fog located on-board of the high-speed train. In particular, local virtual Mobility Management Entity (vMME) will be deployed onboard as part of the EFS to cope with the huge amount of signalling resulted by hundreds of passengers traveling by high-speed train. Also, the deployed onboard units can also be used to host some specific core functions, such as local breakouts, to enable the storage and consumption of content locally, without the need of going through the train’s backhaul connection. Consequently, the burden of passenger’s mobility signalling will reduce significantly on the train’s backhaul.

Conclusion


Edge and Fog are key pillars of the future communication. In the previous work, there is however not yet defined the integrated platform that integrates and federates those two pillars together. In this work, we have presented ongoing work in the related industrial standards, ETSI MEC and OpenFog. Next we presented the 5G-CORAL Architecture, integrated Edge and Fog Computing System (EFS) and the Orchestration and Control System (OCS), Next, the key challenges related to the integration of Edge and Fog were presented categorized in the EFS and OCS respectively. The architecture is based on the ETSI MEC and ETSI NFV frameworks and targets 5G KPI such as increasing data rate, increasing spectrum efficiency, reduction in latency and mobility enhancement. Based on 5G-Coral architecture, several use cases were presented benefit from being hosted in the edge and distributed private fog networks in a near proximity of the end user. Finally, the demonstration testbed platforms were presented to validate the expected solution on top of the 5G-CORAL architecture.

References

[1] ITU-T, “The Tactile Internet”, Technology Watch Report, August 2014.
[2] M. Abrash, “Latency – the sine qua non of AR and VR,” 2012. [Online]. Available: http://blogs.valvesoftware.com/abrash/latency-the-sine-qua-non-of-ar-and-vr/. [Accessed: October 18, 2016].
[3] W. Sun et al. "D2D-based V2V communications with latency and reliability constraints", IEEE Globecom Workshops, Austin, pp. 1414-1419, 2014.
[4] European Telecommunications Standards Institute, “Mobile Edge Computing (MEC); Framework and Reference Architecture,” ETSI GS MEC 003, March 2016.
[5] European Telecommunications Standards Institute, “Network Functions Virtualisation (NFV); Acceleration Technologies; VNF Interfaces Specification,” ETSI GS NFV-IFA 002, April 2016.
[6] OpenFog Consortium, “Mission”, January 2016. [Online]. Available: https://www.openfogconsortium.org/about-us/. [Accessed: October 18, 2016].
[7] OpenFog Consortium, https://www.openfogconsortium.org/
[8] ETSI, “Mobile Edge Computing (MEC); Radio Network Information API,” European Telecommunications Standards Institute, GS MEC 012, July 2017.
[9] ETSI, “Network Functions Virtualisation (NFV); Management and Orchestration; Architectural Options,” European Telecommunications Standards Institute, GS NFV-IFA 009, July 2016.