The “Qiyong Qiantang†2018AI Cloud Eco-International Summit was held in Hangzhou International Expo Center. At the meeting, Hikvision announced the full opening of the AI ​​Cloud architecture and the construction of the AI ​​industry ecosystem with all parties.
What is the concept of the AI ​​Cloud architecture proposed by Hikvision, and what impact does it have on industry applications? At the summit, Dr. Bi Huijuan, Senior Vice President of Hikvision, gave an in-depth explanation.
Human beings have been exploring artificial intelligence for more than 60 years. In recent years, with the breakthrough of deep learning technology, artificial intelligence has begun to enter the stage of industrial application and deep cultivation. A new industrial revolution has begun, and all walks of life are in Launch different AI products and solutions.
AI Cloud is a new architecture for the intelligent upgrade of the Internet of Things industry, aiming to solve many problems facing the current Internet of Things.
Because IoT devices are complex, diverse, and widely distributed, they bring many challenges such as network, computing storage, and operation and maintenance management. Cloud computing is not a panacea in the Internet of Things. The combination of edge computing and cloud computing can better solve the real problem of the Internet of Things. Taking video surveillance as an example, it requires flexible response from edge devices, data and video viewing control of nearby edge devices, and management of complex mining, storage, and computing devices—like complex systems such as human bodies. The connection between the sensory tip and the brain is a very complex meridian system, not a simple information channel.
In the AI ​​Cloud architecture, edge nodes focus on multi-dimensional sensing data collection and front-end intelligent processing; edge domains focus on sensing data aggregation, storage, processing, and intelligent applications; cloud centers focus on business data fusion and big data multi-dimensional analysis applications. Data from the edge node to the edge domain, to achieve "gather edge to domain"; from the edge domain to the cloud center, to achieve "data into the cloud." Domains and cloud centers can be multi-level and multi-class. Depending on the application, the data aggregated in the edge domain and the data transmitted to the cloud will be different in model and content. The role of the marginal domain is like the “midfield†of football. It is responsible for deciding when, what type, and to what extent the data is sent to the cloud center to achieve “on-demand aggregationâ€.
In summary, the cloud-edge integration of Hikvision AI Cloud is not simply a cloud + edge, but it is not simply divided into several so-called "small clouds" and then aggregated into a larger cloud. A series of problems such as application, data processing, and management brought by the cloud. Hikvision AI Cloud follows the core concept of “edge awareness, on-demand aggregation, multi-layered cognition, hierarchical applicationâ€, and designs and builds a three-dimensional intelligent service architecture. In short, AI Cloud is not Cloud CompuTIng, but the sum of infrastructure, data resources, platform services, and applications.
AI Cloud Mission: Four Capabilities + One Ecological GoalIn the field of Internet of Things, AI Cloud will provide four capabilities: arbitrability of AI resources, on-demand convergence of data, scenario response of applications, and integrated operation and maintenance; achieving an ecological goal: a growing AI Cloud ecosystem .
Capability 1: Configurable AI resources. Let AI be like power and become a schedulable resource for information systems. Hikvision has developed a resource management platform in the edge area and cloud center respectively. Taking the "edge domain management scheduling platform" as an example, the platform can manage and schedule the resources of the storage resource pool, the data resource pool, and the algorithm warehouse in the domain, that is, "two pools and one library". At the same time, by establishing an algorithm model specification, multi-vendor algorithms are supported to perform management scheduling in the same algorithm warehouse.
Capability 2: On-demand convergence of data. Let the data flow like water to a controlled place. To achieve accurate transmission and storage, useless data should not be mixed, otherwise the data becomes a "flooding beast" that wastes resources. Taking video as an example, a large amount of raw data is collected from the edge node front-end device, and the data resource management platform is used to clean and use invalid data and generate available valid data. Through the data resource management platform, the available valid data can be accurately flowed to the application. The place can reduce the unnecessary use of network and storage resources, and help to achieve efficient data mining calculations.
Capability 3: The scenario response of the application. Let the app look like sunlight, illuminating every corner of the business scene. If AI and data are raw materials, then the application is the product. Depending on the application scenario, the same raw materials can be combined into different products, and the products can also be compatible and coordinated with each other. The edge domain provides the IoT intelligent application platform, and the cloud center provides a big data integrated application platform, which supports both “most use†and “multiple use of the same numberâ€. For users, the different applications developed by more partners are based on the same platform and meet the different needs of the industry. It is a key appeal, and users can get more than a single application under this architecture. This application development model will be the main mode of industrial application development driven by AI and big data in the future.
Capability 4: Integrated construction of operation and maintenance. Let the operation and maintenance like the air, do not brush the sense of existence but everywhere. Hikvision has developed a unified operation and maintenance service platform that can manage all kinds of equipment, software and application systems of edge nodes, edge domains and cloud centers. The platform provides users with the functions of daily equipment asset management; provides telephone dispatching, automatic inspection and fault location and disposal functions for the operation and maintenance team; provides mobile terminal application functions for maintenance personnel; provides corresponding equipment manufacturers and users Statistical analysis functions, etc.
Goal: Establish a growthable ecosystem. Let the ecology grow like a big tree, rooted in cooperation and fertile soil. Take the public cloud fluorite cloud as an example. It connects tens of millions of terminal devices, and there are tens of millions of active users around the world. It has over 100 million alarm pushes every day and over 10 million video previews every day. At present, Fluorite Cloud has more than 20,000 development users and more than 4,000 active applications. Hikvision is regarded as a stage for partners to build a win-win cooperation and provide an open platform for AI Cloud capabilities. On this platform, equipment manufacturers, basic software vendors, algorithm vendors, data providers, and application developers will find their way into use; Service providers, operation and maintenance service providers and technical specification development teams will find opportunities for cooperation. Through cooperation, we will create two types of ecosystems for products and services, so that more algorithms can be used for users in a unified platform, allowing more teams to easily develop applications that meet the needs of users.
A good ecology is the road to development of Hikvision and its partners, and it is also the road to development of the industry. To this end, Hikvision is fully committed to the platform work, not only provides an open platform for capabilities, but also provides an open experience environment and a compatibility verification environment, laying a solid foundation for the ecology and escorting cooperation and win-win.
PON technology originated from the ATMPON technology standard that began to take shape in 1995. PON is the abbreviation of "passive optical network" in English. GPON (Gigabit-Capable Passive Optical Network) was first proposed by the FSAN organization in September 2002. On this basis, ITU-T completed the formulation of ITU-T G.984.1 and G.984.2 in March 2003. The standardization of G.984.3 was completed in February and June. Thus finally formed the standard family of GPON.
The basic structure of the equipment based on GPON technology is similar to the existing PON. It is also composed of OLT (optical line terminal) at the central office and ONT/ONU (optical network terminal or called optical network unit) at the user side. Optical fiber (SM fiber) and passive splitter (Splitter) composed of ODN (optical distribution network) and network management system.
The technical feature of GPON is that it borrows the GFP (Generic Framing Procedure) technology defined by ITU-T at the second layer, expands support for GEM (General Encapsulation Methods) encapsulation format, and reorganizes services of any type and any rate. It is transmitted by PON, and the GFM frame header contains frame length indicator bytes, which can be used for the transmission of variable-length data packets, which improves the transmission efficiency, so it can support all services more simply, universally and efficiently.
Gpon Olt,4 PON,8 PON,Gpon 16 Ports Olt,10G Port, GPON Device,Uplink Port
Shenzhen GL-COM Technology CO.,LTD. , https://www.szglcom.com