Not only Wi-Fi 6: how Huawei will develop network technologies
At the end of June, the next meeting of IP Club, a community created by Huawei to exchange views and discuss innovations in the field of network technologies, was held. The range of issues raised at it was quite wide: from global industry trends and business challenges facing customers, to specific products and solutions, as well as options for their implementation. At the meeting, experts from the Russian division of corporate solutions and from the company’s headquarters presented their new product strategy for networking solutions, and also opened details about the recently released Huawei products.
Since in the allotted few hours I wanted to accommodate the maximum of useful things, the event turned out to be informative. In order not to abuse Habr's bandwidth and your attention, in a post we will share the main points discussed at the IP Club “river walk”. Feel free to ask questions! We will give short answers right there. Well, those requiring a more thorough approach will be revealed in separate materials.
In the first part of the event, guests listened to reports prepared by Huawei specialists, primarily on the artificial intelligence-based Huawei AI Fabric solution designed to create next-generation super-productive autonomous networks, as well as on Huawei CloudCampus, which promises to accelerate the digital transformation of the business through a new approach. to the organization of cloud computing. A separate unit was a presentation with the nuances of Wi-Fi 6 technology used in our new products.
')
After the conference part, the club members switched to free communication, dinner and observing the beauties of evening Moscow overboard. The general agenda turned out to be something like this - now we turn to specific speeches.
Arthur Wang, head of the IP-direction of Huawei Enterprise in Russia, presented the guests a strategy for developing the company's network products. First and foremost, he outlined a framework based on which the company is adjusting its course in a turbulent market situation (recall, in May 2019, the US authorities included Huawei in the so-called Entity List).
To begin with - a couple of paragraphs on the results achieved. It is not the first year that Huawei invests in strengthening its position in the industry, and invests systematically. Over 15% of the company's revenue is reinvested in R&D. Out of more than 180 thousand employees of Huawei, R&D accounts for more than 80 thousand. Tens of thousands of specialists are engaged in the development of chips, industry standards, algorithms, artificial intelligence systems and other innovative solutions. By the end of 2018, the number of Huawei patents in total exceeded 5100.
Huawei surpasses other telecom vendors in the number of representatives on the Internet Engineering Task Force, or IETF, which is involved in the development of architecture and standards for the Network. 84% of the draft versions of the SRv6 routing standard, which serves as the foundation for building the next-generation 5G networks, were also prepared by Huawei experts. In the Wi-Fi 6 standards development groups, the company's specialists made about 240 proposals - more than any other player on the telecom market. As a result, in 2018, Huawei released the first access point with Wi-Fi 6 support.
One of the main long-lasting advantages of Huawei in the future will be the transition to fully proprietary chips. To bring one ih-house chip made to the market, it takes 3-5 years with an investment of several billion dollars. So, the company started to implement the new strategy in advance and now demonstrates its practical results. For 20 years, Huawei has been improving the Solar series chips, and by 2019, this work culminated in the creation of the Solar S: Esok launches routers for data centers, security gateways, and AR-series enterprise-class routers. As an intermediate result of this strategic plan, a year and a half ago the company released the world's first processor for high-performance routers, designed according to the 7-nanometer process technology.
Another priority of Huawei is the development of its own software and hardware platforms. Including complex VRP (Versatile Routing Platform), which helps to quickly implement new technologies in all product series.
Huawei also relies on the development and testing of new technologies, starting from the integrated product development (IPD) cycle: it allows you to quickly implement new functionality in a wide variety of products. Among the main trump cards of Huawei here is a huge distributed “factory”, with facilities in Nanjing, Beijing, Suzhou and Hangzhou, for automated testing of solutions in the corporate sector. With an area of ​​over 20 thousand square meters. m. and more than 10 thousand allocated for testing ports, the complex allows you to work out over 200 thousand different scenarios for the operation of equipment, covering 90% of the situations that may arise during its operation.
Huawei also focuses on the flexible interaction of parts of its ecosystem, on its own ICT equipment manufacturing capacities, as well as on a cloud service for DemoCloud customers and partners.
But the main thing, we repeat, Huawei is actively working to replace external hardware development in its solutions with its own. The transformation is carried out according to the six sigma management methodology, due to which each process is clearly regulated. As a result, in the foreseeable future, the company's chips will completely replace third-party ones. 108 models of new products based on Huawei hardware will be presented in the second half of 2019. Among them are industrial routers AR6300 and AR6280 with 100GE uplink ports, which will be released in October.
At the same time, Huawei has a reserve in time to make the transition to in-house development: while the American authorities have allowed Broadcom and Intel to supply Huawei chipsets for another two years . During the presentation, Arthur Van hastened to reassure the audience regarding the ARM architecture, which is used, in particular, in the telecom equipment of the AR series: the ARMv8 license (for example, the Kirin 980 processor is built on it) , and by the time the ninth generation ARM processors will enter the scene, Huawei will hone its own development.
Zhao Zhipeng, director of Huawei's campus network division, shared his team's achievements. According to statistics presented by him, Huawei CloudCampus Network Solution, solutions for service-oriented, campus networks, currently serves more than 1.5 thousand companies from large and medium-sized businesses.
As the core of such an infrastructure, Huawei today offers CloudEngine series switches, and first of all, CloudEngine S12700E for organizing non-blocking data transmission over the network. It has a very high switching capacity (57.6 Tbit / s) and the highest (among comparable solutions) port density 100GE. CloudEngine S12700E is also able to support wireless connections of more than 50 thousand users and 10 thousand wireless access points. At the same time, the fully programmable Solar chipset allows you to update services without replacing hardware. Also thanks to it, a smooth evolution of the network is possible - from the traditional routing architecture, which is historically adopted in the data center, to an adaptive network based on software-defined network technology (SDN): a service-oriented network allows gradual development.
In the infrastructure based on CloudEngine switches, wired and wireless networks converge easily: they are managed using a single controller.
In turn, the telemetry system allows you to monitor network devices in real time and visually visualize the activity of each user. And the CampusInsight network analyzer, through the processing of big data, helps to quickly identify possible malfunctions and establish their root causes. AI-based operation and maintenance system greatly reduces the speed of response to problems - sometimes up to several minutes.
One of the main capabilities of the infrastructure with CloudEngine S12700E “at the core” is the deployment of isolated virtual networks for several organizations.
Among the technical innovations that determine the advantages of a network based on CloudEngine S12700E, three stand out:
For their part, King Tsui, chief engineer of the network technology and solutions department of Huawei Enterprise, and Peter Zhang, director of marketing for data center solutions in the same department, presented, for their part, the solutions by which The company helps deploy modern data centers.
Standard Ethernet networks are increasingly providing network bandwidth, which is necessary for modern computing systems and data storage systems. These requirements are growing: according to expert forecasts, by the mid-2020s, autonomous intelligent systems based on more and more sophisticated artificial intelligence and, possibly, using quantum computing will reign in the industry.
There are three main trends in the work of data centers:
To match them, Huawei has created an AI Fabric solution for deploying next-generation networks that can transmit data without loss and with very low latency (at 1 μs). AI Fabric's central idea is the transition from a TCP / IP infrastructure to a RoCE converged network. Such a network provides remote direct memory access (RDMA), moreover, it is compatible with conventional Ethernet and can exist “on top” of the old-style data center network infrastructure.
At the heart of AI Fabric, the industry’s first data center switch based on an AI chip. Its iLossless algorithm optimizes the processes in the network based on the specifics of the traffic and ultimately significantly improves the efficiency of calculations in data centers.
Thanks to three technologies - to accurately identify congestion, to dynamically configure maximum load and for fast backflow control - Huawei AI Fabric reduces latency in the infrastructure, virtually eliminates packet loss and extends network bandwidth. Thus, Huawei AI Fabric is well suited for creating distributed storage systems, AI solutions and high-load computing.
Huawei CloudEngine 16800, equipped with a 400GE network card with 48 ports and an AI-enabled chip and the potential for autonomous infrastructure management, was the industry's first switch with wired artificial intelligence. Due to the analysis system built into CloudEngine 16800 and the FabricInsight centralized network analyzer, it is possible to detect network failures and their causes in seconds. The AI ​​system performance on CloudEngine 16800 reaches 8 TFlops.
Among the main priorities of Huawei is the development of the Wi-Fi 6 standard, which underlies most solutions with a foundation for the future. In his mini-report, Alexander Kobzantsev described in detail why the company relied on 802.11ax. In particular, he explained the advantages of orthogonal frequency division multiple access (OFDMA), which makes a network deterministic, reduces the likelihood of conflicts in it and ensures stable performance even in multiple connections.
Judging by how reluctantly the regulars of the IP Club diverged and what a bunch of questions from them fell on the members of the Huawei team, the meeting was a success. Those who wanted to continue highly concentrated communication about the future of network technologies with like-minded people were interested in where and when the next meeting of the club would take place. True, this information is so secret that even to the organizers it is not yet available. As soon as the time and place of the meeting become known, we will make an announcement.
But what is absolutely certain is that very soon we will write a post on the implementation of CloudCampus with details from our engineers - stay tuned for updates to the Huawei blog. By the way, maybe you yourself want to know something specifically about CloudCampus? Ask in the comments!
Since in the allotted few hours I wanted to accommodate the maximum of useful things, the event turned out to be informative. In order not to abuse Habr's bandwidth and your attention, in a post we will share the main points discussed at the IP Club “river walk”. Feel free to ask questions! We will give short answers right there. Well, those requiring a more thorough approach will be revealed in separate materials.
In the first part of the event, guests listened to reports prepared by Huawei specialists, primarily on the artificial intelligence-based Huawei AI Fabric solution designed to create next-generation super-productive autonomous networks, as well as on Huawei CloudCampus, which promises to accelerate the digital transformation of the business through a new approach. to the organization of cloud computing. A separate unit was a presentation with the nuances of Wi-Fi 6 technology used in our new products.
')
After the conference part, the club members switched to free communication, dinner and observing the beauties of evening Moscow overboard. The general agenda turned out to be something like this - now we turn to specific speeches.
Huawei's strategy: all for one, everything for one
Arthur Wang, head of the IP-direction of Huawei Enterprise in Russia, presented the guests a strategy for developing the company's network products. First and foremost, he outlined a framework based on which the company is adjusting its course in a turbulent market situation (recall, in May 2019, the US authorities included Huawei in the so-called Entity List).
To begin with - a couple of paragraphs on the results achieved. It is not the first year that Huawei invests in strengthening its position in the industry, and invests systematically. Over 15% of the company's revenue is reinvested in R&D. Out of more than 180 thousand employees of Huawei, R&D accounts for more than 80 thousand. Tens of thousands of specialists are engaged in the development of chips, industry standards, algorithms, artificial intelligence systems and other innovative solutions. By the end of 2018, the number of Huawei patents in total exceeded 5100.
Huawei surpasses other telecom vendors in the number of representatives on the Internet Engineering Task Force, or IETF, which is involved in the development of architecture and standards for the Network. 84% of the draft versions of the SRv6 routing standard, which serves as the foundation for building the next-generation 5G networks, were also prepared by Huawei experts. In the Wi-Fi 6 standards development groups, the company's specialists made about 240 proposals - more than any other player on the telecom market. As a result, in 2018, Huawei released the first access point with Wi-Fi 6 support.
One of the main long-lasting advantages of Huawei in the future will be the transition to fully proprietary chips. To bring one ih-house chip made to the market, it takes 3-5 years with an investment of several billion dollars. So, the company started to implement the new strategy in advance and now demonstrates its practical results. For 20 years, Huawei has been improving the Solar series chips, and by 2019, this work culminated in the creation of the Solar S: Esok launches routers for data centers, security gateways, and AR-series enterprise-class routers. As an intermediate result of this strategic plan, a year and a half ago the company released the world's first processor for high-performance routers, designed according to the 7-nanometer process technology.
Another priority of Huawei is the development of its own software and hardware platforms. Including complex VRP (Versatile Routing Platform), which helps to quickly implement new technologies in all product series.
Huawei also relies on the development and testing of new technologies, starting from the integrated product development (IPD) cycle: it allows you to quickly implement new functionality in a wide variety of products. Among the main trump cards of Huawei here is a huge distributed “factory”, with facilities in Nanjing, Beijing, Suzhou and Hangzhou, for automated testing of solutions in the corporate sector. With an area of ​​over 20 thousand square meters. m. and more than 10 thousand allocated for testing ports, the complex allows you to work out over 200 thousand different scenarios for the operation of equipment, covering 90% of the situations that may arise during its operation.
Huawei also focuses on the flexible interaction of parts of its ecosystem, on its own ICT equipment manufacturing capacities, as well as on a cloud service for DemoCloud customers and partners.
But the main thing, we repeat, Huawei is actively working to replace external hardware development in its solutions with its own. The transformation is carried out according to the six sigma management methodology, due to which each process is clearly regulated. As a result, in the foreseeable future, the company's chips will completely replace third-party ones. 108 models of new products based on Huawei hardware will be presented in the second half of 2019. Among them are industrial routers AR6300 and AR6280 with 100GE uplink ports, which will be released in October.
At the same time, Huawei has a reserve in time to make the transition to in-house development: while the American authorities have allowed Broadcom and Intel to supply Huawei chipsets for another two years . During the presentation, Arthur Van hastened to reassure the audience regarding the ARM architecture, which is used, in particular, in the telecom equipment of the AR series: the ARMv8 license (for example, the Kirin 980 processor is built on it) , and by the time the ninth generation ARM processors will enter the scene, Huawei will hone its own development.
Huawei CloudCampus Network Solution - Service Oriented Networks
Zhao Zhipeng, director of Huawei's campus network division, shared his team's achievements. According to statistics presented by him, Huawei CloudCampus Network Solution, solutions for service-oriented, campus networks, currently serves more than 1.5 thousand companies from large and medium-sized businesses.
As the core of such an infrastructure, Huawei today offers CloudEngine series switches, and first of all, CloudEngine S12700E for organizing non-blocking data transmission over the network. It has a very high switching capacity (57.6 Tbit / s) and the highest (among comparable solutions) port density 100GE. CloudEngine S12700E is also able to support wireless connections of more than 50 thousand users and 10 thousand wireless access points. At the same time, the fully programmable Solar chipset allows you to update services without replacing hardware. Also thanks to it, a smooth evolution of the network is possible - from the traditional routing architecture, which is historically adopted in the data center, to an adaptive network based on software-defined network technology (SDN): a service-oriented network allows gradual development.
In the infrastructure based on CloudEngine switches, wired and wireless networks converge easily: they are managed using a single controller.
In turn, the telemetry system allows you to monitor network devices in real time and visually visualize the activity of each user. And the CampusInsight network analyzer, through the processing of big data, helps to quickly identify possible malfunctions and establish their root causes. AI-based operation and maintenance system greatly reduces the speed of response to problems - sometimes up to several minutes.
One of the main capabilities of the infrastructure with CloudEngine S12700E “at the core” is the deployment of isolated virtual networks for several organizations.
Among the technical innovations that determine the advantages of a network based on CloudEngine S12700E, three stand out:
- Dynamic Turbo The technology is based on the concept of “slicing” network resources adopted for 5G networks for various types of traffic. Due to Wi-Fi 6-based hardware solutions and proprietary algorithms, it allows reducing the delay for applications with high network priority up to 10 ms.
- Lossless data transfer. DCB (Data Center Bridging) technology prevents packet loss.
- "Smart antenna." Eliminates “gaps” in the coverage area and is able to expand it by 20%.
Huawei AI Fabric: artificial intelligence in the genome of the network
For their part, King Tsui, chief engineer of the network technology and solutions department of Huawei Enterprise, and Peter Zhang, director of marketing for data center solutions in the same department, presented, for their part, the solutions by which The company helps deploy modern data centers.
Standard Ethernet networks are increasingly providing network bandwidth, which is necessary for modern computing systems and data storage systems. These requirements are growing: according to expert forecasts, by the mid-2020s, autonomous intelligent systems based on more and more sophisticated artificial intelligence and, possibly, using quantum computing will reign in the industry.
There are three main trends in the work of data centers:
- Super-fast transfer of huge data streams. A standard 100 gigabyte switch will not cope with a twentyfold increase in traffic. And today, such a reserve is becoming necessary.
- Automation in the deployment of services and applications.
- Smart O&M. Solving user problems in manual or semi-automatic mode takes hours, which is unacceptably a lot by the standards of 2019, not to mention the near future.
To match them, Huawei has created an AI Fabric solution for deploying next-generation networks that can transmit data without loss and with very low latency (at 1 μs). AI Fabric's central idea is the transition from a TCP / IP infrastructure to a RoCE converged network. Such a network provides remote direct memory access (RDMA), moreover, it is compatible with conventional Ethernet and can exist “on top” of the old-style data center network infrastructure.
At the heart of AI Fabric, the industry’s first data center switch based on an AI chip. Its iLossless algorithm optimizes the processes in the network based on the specifics of the traffic and ultimately significantly improves the efficiency of calculations in data centers.
Thanks to three technologies - to accurately identify congestion, to dynamically configure maximum load and for fast backflow control - Huawei AI Fabric reduces latency in the infrastructure, virtually eliminates packet loss and extends network bandwidth. Thus, Huawei AI Fabric is well suited for creating distributed storage systems, AI solutions and high-load computing.
Huawei CloudEngine 16800, equipped with a 400GE network card with 48 ports and an AI-enabled chip and the potential for autonomous infrastructure management, was the industry's first switch with wired artificial intelligence. Due to the analysis system built into CloudEngine 16800 and the FabricInsight centralized network analyzer, it is possible to detect network failures and their causes in seconds. The AI ​​system performance on CloudEngine 16800 reaches 8 TFlops.
Wi-Fi 6 as the basis for innovation
Among the main priorities of Huawei is the development of the Wi-Fi 6 standard, which underlies most solutions with a foundation for the future. In his mini-report, Alexander Kobzantsev described in detail why the company relied on 802.11ax. In particular, he explained the advantages of orthogonal frequency division multiple access (OFDMA), which makes a network deterministic, reduces the likelihood of conflicts in it and ensures stable performance even in multiple connections.
Conclusion
Judging by how reluctantly the regulars of the IP Club diverged and what a bunch of questions from them fell on the members of the Huawei team, the meeting was a success. Those who wanted to continue highly concentrated communication about the future of network technologies with like-minded people were interested in where and when the next meeting of the club would take place. True, this information is so secret that even to the organizers it is not yet available. As soon as the time and place of the meeting become known, we will make an announcement.
But what is absolutely certain is that very soon we will write a post on the implementation of CloudCampus with details from our engineers - stay tuned for updates to the Huawei blog. By the way, maybe you yourself want to know something specifically about CloudCampus? Ask in the comments!
Source: https://habr.com/ru/post/461413/
All Articles