Future Prospects of an Intent-Driven Campus Network

Today, it is widely recognized in the industry that the future development of campus networks will be largely guided by autonomous driving technologies. As described in Section 2.3, Level 3 autonomous driving networks start to become self-aware and can be context-driven to dynamically adjust and optimize networks based on external environments, implementing intent-based closed-loop management. Level 4 autonomous driving networks will go a step further, unleashing the full potential of digital twins built on campus networks. By that time, networks themselves will be predictable. Upon detecting network parameter changes, networks will take preventive measures before customers can detect problems, avoiding unexpected network incidents.

In the age of connectivity, big data and Artificial Intelligence (AI) technologies will become increasingly mature. As such, networks will further transform towards Level 5 autonomous driving. The biggest difference with Level 5 autonomous driving networks will be that digital twins will not only cover the campus network, but also all things in the entire campus and even in the world at large. Level 5 autonomous driving networks will also run completely independently. This chapter mainly explores the future prospects of Level 4 and Level 5 autonomous driving networks.

INTELLIGENCE VISION FOR FUTURE CAMPUSES

In the age of connectivity, highly promising technologies, such as Internet of Things (IoT), cloud computing, ultra-broadband, big data, and AI, will become mature. Powered by these technologies, campuses will become fully digitalized, and all things in campuses will generate data. The data will represent the things, statuses of these things, and even various manually-defined concepts.

With ultra-broadband technology, no network bandwidth bottlenecks will exist, and all data will be migrated to the cloud and participate in computing in real time. In addition, with assistance from the super computing power of cloud computing and various mathematical analytics models, AI systems will continuously mine the complex relationships between things represented by big data to achieve a revolutionary leap in cognitive recognition of our world.

In future campuses, the information silos that arise from separately constructed networks on traditional campuses will disappear. All service data will be fully shared, achieving full convergence between the physical and digital worlds. Ultimately, a digital model for the entire campus will be built and updated in real time. By that time, the campus will realize a closed-loop process, starting from sensing the touch points of the physical world, to making decisions in the digital world, and then back to intelligently taking action in the physical world. This process will be fully autonomous, and human participation will not be necessary.

As shown in Figure 13.1, a digitalized campus will be similar to an organic life form. That is, the brain of the organic life form will be an AI

controller, its nervous system will be the physical network of the campus, and its nerve endings will be various service terminals and the digital systems inside and outside the campus.

NETWORK TECHNOLOGY PROSPECTS FOR FUTURE CAMPUSES

FOR FUTURE CAMPUSES_

A network’s most important aim is to ensure the secure, fast, and accurate transmission of information. An autonomous driving network can effectively ensure the attainment of this goal due to its predictive capabilities. This is particularly true for Level 5 autonomous driving networks that will act upon the big data-based prediction results across the entire campus. On Level 5 autonomous driving networks, all things will be centrally controlled and managed; traffic will be adjusted before network congestion occurs; preventative measures will be taken before network faults occur; and network capacity will be automatically expanded due to insufficient network performance. These measures all ensure the secure, fast, and accurate transmission of information, while, more importantly, also improving network Operations and Maintenance (O&M) efficiency and reducing network deployment and O&M costs.

A Level 5 autonomous driving network, however, can only be achieved through meticulous design. Firstly, we should create an edge intelligence layer that can sense environments in real time on the physical network and greatly simplify network architecture and protocols. Secondly, we need to build a digital twin network through unified modeling to ensure the traceability and predictability of the global situation, and also implement predictive O&M and proactive closed-loop optimization using AI technologies. Lastly, we must establish an open cloud platform to achieve AI algorithm training and optimization, enable the agile development of various applications (covering planning, design, service provisioning, O&M assurance, and network optimization), and facilitate automatic closed-loop operations throughout the whole lifecycle.

1. Intelligent and simplified campus network architecture

A campus network typically involves three parts: the control system, control mode, and physical network.

Control system: On a Level 4 autonomous driving network, the previously independent Software-Defined Networking (SDN) controller, Wide Area Network (WAN) controller, and Data Communication

Network controller on Level 3 autonomous driving networks become fully collaborative. That is, the entire campus network will use only one unified AI controller.

As AI technologies become mature, we will most certainly embrace Level 5 autonomous driving networks. On a Level 5 autonomous driving network, the AI controller on the campus network will fully collaborate with all other AI controllers on the campus (for example, those controllers for environment monitoring, logistics, warehousing, and production management). By then, only one unified AI controller will be used throughout the entire campus. This AI controller will even detect data outside the campus upon authorization to better steer campus operations and improve user experience.

Control mode: On a Level 4 autonomous driving network, administrators will no longer need to manage the specific configuration logics and commands of network functions. Instead, they will only need to express their intent to the network, and the network will intelligently adjust itself based on the administrator’s intent. When it comes to a Level 5 autonomous driving network, the AI controller will automatically identify administrators’ intent, and administrators will be only required to confirm key decisions.

For example, let us say ten guests are planning to come to your city tomorrow by plane, visit your campus, and attend a conference. In this example, Level 4 autonomous driving network administrators will only need to inform the network of the guests’ plans on the campus. The AI controller will then be able to automatically determine the network service levels and security levels required for these 10 guests, and deliver related configurations to each network device. The entire process will not require network administrators to know about the different service levels, security levels, and network feature configuration logics implemented.

However, the control mode on Level 4 autonomous driving networks will not be completely without manual intervention; rather, manual intervention will be required if unexpected changes affect the network. For example, in the above example, if the number of guests changes on the day when these guests are expected to visit, network administrators will have to manually submit these changes on the network for adjustment. This is because on a Level 4 autonomous driving network, only the campus network will have achieved control collaboration.

However, the preceding case will be totally different on a Level 5 autonomous driving network. Specifically, the campus network control system will automatically identify and track information changes and adjust network resources accordingly. Guests will only be required to authorize their travel information to the campus AI controller. The campus AI controller will automatically arrange pickup vehicles based on the updated flight information, weather information, and road traffic status, and adjust campus network resources such as Wi-Fi according to the updated guest arrival time. The entire process will not require manual follow-up and intervention. Instead, network administrators and guest reception personnel will only act according to the guest itinerary information changes provided by the campus AI controller.

Physical network: Currently, we construct service networks separately to ensure convenient O&M and clear responsibilities. For example, we build separate security networks (e.g., video surveillance and fire protection networks), office networks, and production networks. However, with the rise of virtual network technologies such as Virtual Extensible Local Area Network (VXLAN), we can build one single physical network on an intent-driven campus network.

On a Level 5 autonomous driving network, the campus AI controller will take over all manual O&M tasks and automatically detect information such as campus event plans as well as logistics and warehousing data. By that time, terminals on the physical network will not be limited to traditional network terminals such as computers, printers, cameras, and mobile phones, but will also include IoT terminals and various sensors such as curtain sensors, light sensors, wind sensors, robots, and drones. If the physical network cannot support these upcoming services, the campus AI controller will automatically adjust the physical network devices and topology, including purchasing and leasing new devices. In a similar way, if the campus AI controller detects that spare network devices are overstocked in the warehouse, the campus AI controller will sell or lease out devices to effectively control the physical network’s construction and O&M costs. In addition, future network devices will most likely become modularized. That is, network devices will become similar to desktop computers, with each component being flexibly removed and replaced. In this case, the campus AI controller will manage each component in a refined manner.

2. Intelligent and simplified campus network design, deployment, and O&M

In the future, the campus network architecture will become more intelligent and simplified. This, in turn, will greatly reduce the required investment and costs in the design, deployment, and O&M phases of the campus network.

A Level 4 autonomous driving network will independently design the campus network based on the requirements and related information provided by technical personnel. However, manual adjustment will still be required before network deployment, and manual configuration and commissioning will also be required during network deployment. However, on a Level 5 autonomous driving network, enterprises will only need to set their budget range. Then, the AI controller will automatically determine the most appropriate network architecture and select the optimal devices for the campus. The AI controller will also arrange robots to complete site surveying as well as device configuration and commissioning.

The following takes an enterprise that intends to create a branch in city A as an example. After the enterprise sets its budget range, the AI controller automatically collects and analyzes all information about the branch. The collected information would include the enterprise’s internal information (for example, business and staffing planning for the branch’s next 10 years), the climate information of city A, topographical information of the branch, public security situation of city A, logistics and transportation information of city A, inventory material information of the enterprise, and recently purchased material information. Based on this comprehensive information, the AI controller will automatically produce multiple campus network design solutions for the enterprise, with detailed descriptions of the differences between the different solutions. The designed solutions will not only involve office and production network services, but also all possible digital services in the campus, covering living and entertainment, security defense, and environment ecosystem construction. What’s more, the AI controller will automatically generate physical network plans (for example, bandwidth planning and Wi-Fi network management) based on future service planning, mobile office requirements, and IoT requirements.

Once an enterprise finalizes its design solution, the AI controller will start to deploy the campus network based on the finalized design solution. Specifically, the AI controller automatically purchases goods, tracks goods delivery and logistics, and arranges robots for site survey, civil engineering, and cabling, as well as device installation and commissioning. Network administrators can view the detailed data and reports of the entire process, and they only need to adjust the solution upon any force majeure events. After deployment, the campus network is continuously optimized and expanded based on the network’s actual service running status. Service adjustment is linked with the IT system to ensure that when the IT system detects new service requirements, it automatically triggers the AI controller to provision new services.

The following uses today’s Wi-Fi network optimization and management pain points as an example to envision the future design and deployment process. Strictly speaking, a Wi-Fi network is a self-interference system. To ensure best results, Access Point (AP) interference in the campus should be constantly detected and then continuous optimizations should be carried out accordingly. The current solution involves importing the building drawings of each area in the campus to the software and then using the software to simulate the signal strength of APs. Then, the AP installation position is determined using the software. After that, personnel are assigned to perform a site survey to evaluate the impact of software installation, environment, and electromagnetic interference on Wi-Fi signals, and then determine the final locations of APs one by one based on personal experience. After APs are installed and go online, we need to manually check, perform acceptance tests on, and optimize AP signals. In subsequent network O&M, Wi-Fi network optimization or reconstruction affects many aspects. For this reason, network administrators do not optimize or reconstruct Wi-Fi networks in an area until users submit complaints about network quality. This will no longer happen on Level 5 autonomous driving networks.

On a Level 5 autonomous driving network, the Wi-Fi network design will not be based on the personal experience of an engineer or the model recommended by an expert. Instead, an optimal design solution will be automatically obtained through big data analytics by the AI controller that features powerful computing and learning capabilities. A wireless interference avoidance mechanism will be designed for all things that may cause interference on the entire campus, rather than only focusing on network devices. In addition, the location of wireless terminals will depend on the simulation data model obtained after the entire campus is constructed, rather than on campus construction drawings.

In the deployment phase, robots will perform a site survey and adjust the digital campus model in real time. Then, based on the optimal construction sequence provided by the AI controller, robots will directly perform campus network deployment and related engineering tasks. During the construction phase, the digital campus model will also be continuously updated to ensure that the network’s construction is aligned with the design. Then, once deployment is complete, the Level 5 autonomous driving network will use probe data to simulate the running of real service data on the Wi-Fi network. Meanwhile, the AI controller will collect, analyze, and evaluate the probe data.

In the daytime, the network will constantly self-diagnose and self- analyze itself; meanwhile, in the evening, the network will self-opti- mize and self-adjust as well as deploy and debug new policies. For example, if network bandwidth is insufficient, the AI controller will automatically adjust network bandwidth, policy, and architecture based on the company’s budget, actual number of users and devices, and real campus environment. If capacity expansion is required, the AI controller will check the inventory and logistics capabilities, automatically determine the cost-effectiveness of each capacity expansion solution, and provide the optimal capacity expansion solution for enterprise managers. Then, with the manager’s approval, network capacity will be automatically expanded.

In the O&M phase, administrators on a Level 3 autonomous driving network need to manually operate the SDN controller’s Graphical User Interfaces (GUIs) to detect and locate faults. However, on a Level 5 autonomous driving network, the network will analyze user service experience in real time. When service experience deteriorates or potential faults are predicted, the network will record the anomalies, analyze their root causes, and optimize the network in real time. Once optimization is complete, the network will continuously follow up the optimization results until user experience returns to normal. What is more? The entire process will not require any manual intervention.

If a network fault occurs without any precursor, the network will automatically analyze information such as logs and alarms generated along with the fault, and rectify the fault in real time. All related spare parts will be sent to the fault point immediately. Then, once the consumption of spare parts is confirmed, the warehouse will automatically purchase and supplement spare parts by itself.

Let us use frame freezing in a video conference as an example. Generally speaking, frame freezing is caused by packet loss. With regards to autonomous driving networks at lower levels, after receiving a frame freezing fault report from a user, the administrator then has to find the packet loss point based on the source and destination. However, this fault locating method is inefficient, because it is difficult to reproduce the frame freezing fault.

This situation is entirely different on autonomous driving networks at higher levels. Specifically, the AI controller collects all traffic on the network in real time, analyzes the traffic types, and determines the flow quality in real time for each traffic type. If the quality of a flow is poor, the diagnosis mechanism is automatically triggered to locate the packet loss point. Then, once the packet loss point is found, the AI controller automatically analyzes the running status and logs of the packet loss point and performs optimization accordingly. Following optimization, the AI controller continues to verify whether the corresponding service experience has been improved and then decides whether to further optimize. In this way, fault diagnosis is changed from passive to proactive, and fault rectification is shifted from post-event to real-time, effectively enhancing the satisfaction of campus network users.

The information that can be sensed and referenced by the network will not be limited to the information of the network itself. Therefore, the success rate of network fault prediction and defense will be greatly improved, and services will run uninterrupted due to reliability assurance that has been planned during the network design phase.

3. Intelligent and simplified campus network security

After a campus goes digital, security-related services still remain the top priority. Before the arrival of Level 4 autonomous driving networks, network security is primarily implemented by deploying security components such as firewalls on security borders to prevent network attacks. Such security components fend off network attacks by relying on a sound antivirus database. However, these security components are ineffective towards new security threats.

On a Level 4 autonomous driving network, security will be implemented by focusing on security protection. That is, the network will be capable of real-time threat awareness. In addition to building a security protection system with the global virus database and attack modes, the network will also monitor and analyze traffic behaviors in real time and determine whether to take preventative security actions based on attack characteristics. This approach will not only detect known security threats, but also effectively identify unknown security threats.

In terms of security, a Level 5 autonomous driving network will demonstrate even further progress. The network will predict possible attacks based on the data of all people, incidents, things, and environments in the campus in real time, and even simulate to automatically perform attack-defense drills on the campus data model.

For example, after predicting that a typhoon is about to land, the AI controller will use a campus digital model to simulate the possible impacts of the typhoon on the campus and then update data in real time based on typhoon tracking and wind speed changes. If it determines that the typhoon will cause damage to houses or articles, the AI controller will assign robots to perform reinforcement and apply for a budget to purchase spare parts to be used once the typhoon passes. When someone passes through a dangerous area, the AI controller will remind them to minimize the impact of force majeure factors such as typhoons on the campus.

4. Intelligent and simplified campus network team

A digital campus requires enterprise Information and Communication Technology (ICT) teams to shift their focus from hardware features and performance to the overall solution. This is truly a radical change that will pose higher requirements on products and services from ICT vendors, as well as lead to deeper collaboration between ICT teams while most likely combining the network, application, and security teams.

When purchasing products for a digital campus, enterprises will not simply evaluate vendors based on function fulfillment or performance indicators. Instead, they will evaluate the end-to-end problem-solving capabilities of vendors’ solutions according to their service scenarios. As such, vendors will need to have a deeper understanding of enterprise services. Only then can they perform scenario-specific abstraction, which can be converted into a simple, easy-to-use man-machine interface. And in doing so, they can truly help enterprises improve production efficiency internally and enhance customer experience externally.

In the future, Huawei will lead industry development in five directions: redefining the technical architecture, reshaping the product architecture, setting the industry pace, resetting the industry direction, and opening up new industry space. Huawei will also strive to break the limits in four aspects to create a better future:

a. Redefine the Moore’s Law and challenge the Shannon limit to build the best connections in the world

b. Reshape the computing architecture to make computing power greater and more economical

c. Create the best hybrid cloud to enable industry digitalization

d. Enable full-stack, all-scenario AI for ubiquitous intelligence

Partners and developers are the key to AI technology development. Considering this, Huawei released the Developer Enablement Plan and Shining Star Plan in 2018, which provide solid support in terms of resources, platforms, training courses, and joint solutions, thereby building a foundation for partner applications. The road to autonomous driving networks will be a long one for the ICT industry. Given this, all parties in the industry should work together to move forward in a common direction. Indeed, Huawei is committed to providing leading ICT solutions through continuous innovation, taking on complexity while creating simplicity for customers, and embracing a fully connected, intelligent world together with enterprises and partners.

 
Source
< Prev   CONTENTS   Source   Next >