Disclaimer: This article is a blogger original article, follow the CC 4.0 BY-SA copyright agreement, reproduced, please attach the original source link and this statement.
Calculating an edge (Edge Computing) extending to the edge of the cloud, edge calculation herein, fog calculation carried out MEC, Cloudlet, distributed computing cloud edges and other concepts and definitions art, architecture, and other scenes comparative analysis, and given the forecast and prospects of the development trend of technology.
One
Outline
In the context of the industry's digital transformation, in IoT, 5G, VR, AI and other services under a cloud of demand-driven and technology to promote development, edge computing (Edge Computing) concept came into being and quickly got the attention of the industry.
With respect to the classical cloud computing to give the "cloud" of massive computing power to achieve the sink edge computing resources and services to the edge position, thereby reducing interaction delay, reduce the network load, rich business types, process optimization services, improve quality of service and user experience.
There is no clear definition of the edge of computing concepts, fog calculation, MEC, Cloudlet, edge computing, distributed one after another cloud concepts, ETSI, ITU, OpenFog, ECC, OEC, 3GPP, ISO, IEC, IEEE, Linux Foundation, Foundation for the OpenStack and other industry mainstream standardized, open source and industry organizations are actively promoting but has the focus. In this paper, the edge computing, fog calculations, defined within the MEC, Cloudlet, distributed in areas such as cloud core technology, architecture, scene, etc. Introduction and comparative analysis.
two
Fog computing
Fog computing (Fog Computing) is a concept proposed by cisco 2011, OpenFog Union is the main promoter of fog calculation, the fog calculation is defined as:
a system-level architecture, computing, storage, networking, control and decision-making resources and distribution services to any location from the cloud to the matter, to address the IoT, AI, VR, 5G and other business scenarios demand.
① level architecture: Support for multiple vertical industry applications will be distributed intelligent and services to users and business; ② to cloud service was continuous: the distribution of services and applications between the cloud and was closer to the position of the object; ③ system level: the whole system between the object and the clouds, and then from the cloud to the network edge thereof, a plurality of protocol layers covering, or not part of a specific end system protocol.
Haze and cloud computing interdependent, complementary;
Some features adapted to perform the calculation by the fog nodes, some features are adapted to run on a cloud; DETAILED boundary depending on the particular application scenarios and different network environment. FCN (Fog Computing Node, fog computing node) access network is an intermediary between the intelligent terminal and the intelligent network cloud element, may be physical or virtual, and with the intelligent terminal equipment or access network tight coupling, providing data management and communications service functions between the terminal device and the cloud.
OpenFog Union given haze calculated reference architecture shown in Figure 1: wherein, described primarily from the mist computing nodes, software, angle requirements:
FIG mist 1-OpenFog calculated reference architecture Alliance
(1) Haze computing node:
main components including resource node, the node management, a protocol abstraction layer. Wherein the node comprises a computing resource, storage, networking, accelerators; management node includes hardware configuration and security management; adaptation protocol implemented abstraction layer and the docking sensor terminal device supporting heterogeneous terminals, interworking between a compatible node fog .
(2) Software:
it includes three layer node management and software backplane, application support and application services. Wherein the node manager is responsible for hardware and software configuration of the system nodes or fog and state maintenance, and maintain their availability, reliability, and performance; software running node backplane various types of software, and realizes communication between nodes; providing application management support application, running when the engine, application server, news and events, security services, data storage and management, analytical tools and frameworks for more software applications (micro-services) use and share; application services based on deployment scenarios, resource availability varies , including connectivity services, support services, analysis services, integration services, UI services.
(3) functional requirements:
including performance and scalability, security, manageability, data analysis and control, IT services and applications across nodes. Comprising: a scene matching in accordance with the demand required traffic model, to ensure non-interfering between nodes separated from each other; virtualization technology and by means of lifting containers scalability; providing node security, data security, network security, security monitoring and management from the cloud to the network the whole system was then safety edge; may include management of management interfaces, nodes fog life cycle management, independent of management and the like; and massive data collection, storage, transmission, analysis and other functions to sink to the network edge, close to the data source , needed to achieve a particular data processing capabilities; implement applications and services deployed on-demand computing hierarchy in fog and interoperability to support data sharing between nodes, support applications across nodes.
Fog can be used to calculate IoT, 5G, AI and other scenes, solve localization security, customers location-aware, flexible deployment and scalability, low latency requirements, etc.. Which, IoT is the focus for FC scenes, transportation, smart cities, smart buildings, industrial manufacturing, retail, health care, agriculture, government forces, the wisdom of the family, operators, etc. are applied. Some typical scenarios such as intelligent cars and traffic control, visualization, security and surveillance, smart city and so on.
three
GUY
MEC is the ETSI concept and the main push, experienced the evolution computing (Mobile Edge Computing) moves from the edge to the edge of the multi-access computing (Multi-access Edge Computing) is. Wherein
the mobile computing is to provide an edge IT services in cloud computing environment and the mobile network edge, the network traffic will sink closer to the user's wireless mobile access network side, designed to reduce latency, and efficient network traffic distribution control improve the user experience; multi-access edge computing refers to the provision of IT services and cloud computing capabilities to application development and content providers at the network edge, the environment for the application to provide ultra low-latency, high-bandwidth, real-time access, etc. ability.
MEC Key features include: the nearest access, low delay, visible location, data analysis. ETSI MEC analysis system reference architecture shown in FIG. Which includes the following components:
FIG 2-ETSI MEC System Reference Architecture
(1) MEC Host:
consists of virtual infrastructure and MEP, MEC for carrying all kinds of applications. Wherein the virtual data plane infrastructure responsible for enforcing traffic rules received by the mobile edge of the platform, and forward traffic. MEP (MEC Platform, MEC platform) offers a range of features to achieve the MEC application on a specific virtual infrastructure running and providing mobile edge services.
(2) MEC application:
is a virtual machine running on the MEC host virtualization infrastructure support platform to interact with the MEC, MEC to build and provide services. MEC application with a set of rules and requirements, resources, maximum delay, and other available services include needs. These requirements were confirmed by the MEC system-level management.
(3) Mobile Edge host-level management:
including the Mobile Edge platform manager MEPM, virtualized infrastructure manager VIM, the mobile edge hosts and applications running on it to manage. Which, VIM (Virtualisation infrastructure manager, virtualized infrastructure manager) functions can reference ETSI NFV VIM, the main provider of virtualized resources management functions; MEPM (Mobile edge platform manager, Mobile Edge platform manager) main achievement of Application Lifecycle Management , MEP network element management, application rules and requirements management.
(4) moving the edge level management system:
including MEO, OSS and other components. MEO (Mobile edge orchestrator, moving edge formatter) is a core component, responsible for maintaining the movement of the edge of the overall system view, activating an application package, select suitable mobile edge based hosts Constraints for applications instantiated trigger the application examples and termination, trigger the application re-positioning. According to business needs and solution architecture, ETSI MEC scene will be divided into three categories, each type of solution architecture convergence, vary between categories:
① customer-facing services, including gaming, remote desktop applications, enhancements and auxiliary reality, aid cognition.
② operators and third-party services, including the activities of location tracking devices, big data, security, and other business services.
③ QoE and improve network performance, including the content and DNS cache, performance optimization, video optimization.
Among them, including seven scene: intelligent mobile video acceleration, surveillance video flow analysis, AR (Augmented Reality), intensive computing assistance, corporate network, car network, IoT gateway.
four
Cloudlet
Cloudlet concept 2013 Carnegie Mellon University (Carnegie Mellon University) proposed, from mobile computing, IoT and cloud computing integration, stands for "mobile / IoT devices -cloudlet- cloud" three-tier architecture of the middle layer , it can be seen as a "data center in a box", aimed at making cloud closer to the user.
Based on this, the definition given edge OEC calculated: edge computation to provide a compact data centers close to the user side (edge node), to enhance the user experience of computing and storage resources.
Cloudlet include four characteristics: only software deployment form comprising computing / connection / security capability nearest deployed standards-based cloud constructed like.
Computing edge OEC given reference architecture shown in Figure 3, includes a mobile device, the edge server, the backend system components. Wherein: Cloudlet implemented primarily based on edge server, comprising three layers:
(1) infrastructure layer: hardware, virtualization and management;
(2) open cloud platform: provide operational support app environment and the ability to open, and unified management;
(3) Application: Application of carrying various types of mobile devices unloaded (FIG. 3) based on the virtual machine instance.
FIG edge yl 3-OEC computing system reference architecture Cloudlet
Cloudlet and OEC primarily for integration with mobile cloud computing scene for four types of scenarios, including cloud service highly responsive, extended edge analysis, privacy policy enforcement, cloud masking interrupts. Among them, the typical scenarios include: VR, vCPE, business services (such as virtual desktops), public security, sensor data services, automotive services, mobile app optimization, Industry 4.0, UAV support services, health and sports services, online games, communication service optimization.
Fives
Edge computing
ECC (EdgeComputing Consortium, edge computing industry alliance) was established in 2016, is an active promoter edge computing.
ECC edge computing definition: at the network edge side close to the object or data sources, converged network, compute, storage, distributed open platform application core competencies, the nearest to provide edge intelligence services to meet the industry digitized quick connection, real-time business, data optimization, application key demand of intelligence, security and privacy protection.
It can serve as a bridge link the physical and digital worlds, enabling intelligent asset, intelligent gateway, intelligent systems and intelligent services.
ECC believes edge computing and cloud computing are two important support digital transformation of the industry, both in collaborative networks, services, applications, and other aspects of intelligence will contribute to the wider industry support digital transformation scene and create greater value . Wherein, cloud applicable to non-real time, a long period data, business decisions scenario, while the edge is calculated in real-time, short-period data, local decision making aspects of the scene have an irreplaceable role. Key features include coupling EC given ECC, data first inlet, binding, distribution, integration and the like.
ECC
edge computing reference architecture proposed as shown in FIG.
Wherein the hierarchy includes the following components:
FIG edges 4-ECC calculated reference architecture
(1) ECN (Edge Computing Node , an edge node calculation):
the infrastructure layer, a virtual layer, constituting the edge of the virtual service, providing a bus protocol adaptation, data flow analysis, sequence databases, and other common security services, and demand integrated industry-specific applications and services.
(2) join calculated Fabric:
a coupling and virtualization of computing services layer, shielding ECN heterogeneous nodes, providing resource discovery and scheduling, supporting data and knowledge sharing among ECN model nodes to support the traffic load dynamic scheduling and optimization, support for distributed decision-making and policy enforcement.
(3) Business Fabric:
modeling workflow, from various types of functional services according to a certain logic composition and cooperation, support for the definition of workflow and workload, visual rendering, semantic checks and policy conflict checking, Business Fabric, services, etc. model version management.
(4) Intelligent Services:
Development Services Framework integrated with an integrated development platform and tool chain edge computing and vertical industry model library to provide full life cycle service model and application; deployment of operational services mainly to provide business orchestration, application deployment and application of market and other three item core services.
(5) Service Management:
Support for unified management of endpoint, network, server, storage, data, application isolation, security, distributed architecture; engineering support for integration, deployment, services and data migration, integration testing, validation and integration acceptance lifecycle management.
(6) Data Lifecycle Services:
providing data pre-processing, data analysis, data distribution and policy enforcement, data visualization and storage services. Fabric support the business logic defined by the business lifecycle of data, real-time to meet business requirements.
(7) security services:
include node security, network security, data security, application security, security situational awareness, identity and authentication management services, covering all levels of edge computing architecture, and offer different features for different security levels on demand .
Edge computing by combining industry with the use of scenarios and related applications, according to the characteristics and needs of different industries, complete solutions from a horizontal platform to a vertical landing industry, built many innovative vertical industry solutions in different industries. Currently, ECC core scenario given mainly for IoT, examples include: networking ladder, smart water, smart buildings, smart lighting.
six
Distributed Cloud
Distributed cloud concept ITU-T proposed in 2016 and the main push is an extension of the cloud computing model, in both the cloud computing network-centric, to serve as a way to provide highly reliable pool of resources and transparency, high scalability features at the same time, based on business / user requirements, flexible, agile, on-demand, intelligent delivery of distributed, low-latency, high-performance, secure, reliable, green energy, the ability to open the information technology infrastructure to meet the various sectors of the digital society as a whole transition needs.
Distributed cloud has the following main characteristics capability: a distributed, low-latency, high-performance, secure, reliable, green energy, the ability to open and so on.
Distributed cloud system framework shown in FIG. Distributed cloud system includes two types of components:
FIG 5-ITU-T framework distributed cloud
(1) distributed cloud nodes:
it is an independent self-node cloud management capabilities, to provide cloud infrastructure and services, provides the operating environment for all types of services and applications; based on location, size and node in a distributed cloud system the role of the core into the cloud, the cloud region, such as different edge cloud cloud nodes; according to the needs of different business scenarios, or adopt a more flexible combination of nodes, deploy service on demand capability.
Core Cloud node:
based on large-scale cloud infrastructure to support cloud services to provide a complete system including IaaS, PaaS, SaaS, providing support and operating environment for all types of distributed cloud services and applications with massive data analysis and processing capabilities.
Area cloud node:
functional architectures and cores for cloud similar to the large-scale cloud infrastructure, providing a regional focus on cloud infrastructure and services, providing support and operating environment for regional services and applications on demand to meet business and regional security , regulations and governance requirements.
Cloud edge node:
at the network edge, close to the user and the data source location, to provide local infrastructure resources, data processing, edge applications and services. For different business scenarios, a different edge node cloud deployment location, shape, capacity, different functional requirements.
(2) distributed cloud management system:
unified management of distributed cloud nodes, including unified resource management, network elements and application management, service scheduling, operations management, service management, security management, R & D support, systems integration management, data management.
Focused on the cloud distributed cloud between nodes, and the synergy between the edges of the core, to provide a richer than the calculated position of the single operational capabilities and applications for a variety of business scenario, which includes both the classic cloud scenes, but also including edge computing scenarios, also including edge, core collaboration scenarios, covering 5G, IoT, AI, safety, CORD, CDN, cloud services and other business scenarios.
Seven
Analysis of core technology
Fog calculation, MEC, Cloudlet, edge computation, and other technologies are distributed in the cloud business needs and new drive technology, integration of the "edge" of a specific implementation of the concept of cloud computing, as briefly shown in Table 1 Analysis where the main difference is:
(1) Positioning:
Fog aims to build a computing system, cloud services connected to the object, and cloud collaborative operation; MEC is considered 5G key technologies; Cloudlet is deployed miniature nearby data center; edge computing aims to build open platform edge; then distributed cloud core classic cloud, cloud area and cloud edge of unity as a whole.
(2) Scenario:
Fog calculate the focus for IoT; MEC RAN main provider of mobile application solutions; Cloudlet achieve computing capabilities unloading at the network edge; edge computing mainly for IoT intelligence services; distributed cloud is intended to cover a variety of scenarios.
(3) application and data:
mist calculation, MEC, Cloudlet focus on providing support environment enabling operators and application-oriented, manage and automate the deployment applications providing services; cloud edges and distributed computing application in addition to support, also providing diverse data collection, storage, processing, analysis, policy enforcement capabilities, including hierarchical coordination system constructed localized data processing and data processing Drive large (as shown in table 1).
Eight
Outlook
Overall, the edge computing is still in its infancy, the development is located in the Gartner hype peak on the maturity curve, the future will usher in the explosive growth.
Edge computing technologies include core edge nodes (e.g., routers, switches, base stations, vCPE, data centers, etc.), edge network, edge management system, applications, and services edge. Wherein the hardware infrastructure and software technology have been mostly involved, but the need for adaptation or optimization calculation based on the edge requirements such as support for the edge node performance computing power, edge node, reliability and disaster recovery, calculated edge unified management of intelligent scheduling, heterogeneous edge node tasks, data distribution mechanisms and consistency, edge computing network architecture and performance optimization, large-scale applications and services edge, edge features and technologies (such as data refinement, video compression and analysis ), all of which needs further study.
Author
Introduction
Chen Tian,
now working in Guangdong, China Academy of Telecommunications SA, the main research directions for cloud computing, SDN / NFV.
Chen Nan,
Ph.D., now at the Chinese Academy of Telecommunications Co., Ltd. of Guangdong, the main research directions for the Internet, cloud computing, IDC.
Lee Chun,
Ph.D., now at the Chinese Academy of Telecommunications Co., Ltd. of Guangdong, the main research direction for the data network, IPv6, SDN / NFV.
Fan Yongbing,
Ph.D., senior member of China Institute of Communications, Senior Engineer, China Telecom IP talent, the current global cloud network node OpenCirrus person responsible for China Telecom, China Telecom Laboratory - head of the Laboratory of multimedia data, China Telecom data center networks supporting major members engaged in Internet research, currently responsible for China Telecom cloud computing, IDC and so on.
references
1 OpenFog Consortium,OpenFog Reference Architecture for Fog Computing, February 2017
2 NIST SpecialPublication 800-191 (Draft), The NIST Definition of Fog Computing, August 2017
3 ETSIGS MEC 003 V1.1.1 (2016-03), Mobile Edge Computing (MEC); Framework andReference Architecture
4 OpenEdge Computing, Open Edge Computing- From Vision to Reality, 21 June 2016
5 EdgeComputing Consortium, Edge Computing Reference Architecture 2.0, November 2017
6 ITU-T， Draft Recommendation ITU-T Y.ccdc-reqts:Distributed cloud overview and high-level requirements
Forwarding proliferation
Together to promote the dissemination of knowledge edge computing
Recommended Reading