New Generation of WAN Solution—ADWAN

    10-10-2022
Overview

WAN is mainly responsible for the long-distance interconnection between headquarters and branches, branches and branches, data centers and data centers, and so on. Currently, WAN is mainly passively "bearing" the traffic of various services. With the development of cloud computing, mobile Internet and other application models and changes in traffic models, the network needs to actively "adapt" to business traffic of applications, so that services can change as needed. Due to the fact that current network management is primarily device-oriented, and active O&M capabilities from a business perspective are insufficient, complex network services, difficult upgrade and expansion, and closed and rigid architecture have become increasingly prominent issues. Traditional distributed networks focus on the performance and O&M of a single network node, but lack business scheduling and optimization from a global perspective based on in-depth business perception.

The Application-driven Wide Area Network (AD-WAN) solution is a standard, converged, layered, open and smart SDN network architecture. The intelligent big data analysis, SDN controller, and network management are converged in a unified way, to achieve a converged network control center and network brain that integrates management, control, and analysis, for network-wide coverage, and end-to-end service arrangement. It is user-centric with unified Portal, to truly fulfill "one login, issuance with one key, overall protection, one-station O&M". Underpinned by the big data analysis and AI learning capabilities, the real-time network snapshots, offline modeling are captured to realize intelligent network insight, intelligent simulation, and intelligent troubleshooting. Through the support of SRv6, the capabilities of cloudification and network programming are essential. The AD-WAN solution helps enterprise users realize the intelligent upgrade of digital networks.

The above figure shows the overall architecture of the AD-WAN solution, which consists of the network device layer, control layer, and management arrangement layer.

* Network device layer: Network devices are controlled and managed by the SDN controller, providing extensive standard protocols to connect with the control layer, such as SNMP, NETCONF, Telemetry, BGP-LS, BGP Flowspec, and PCEP. Data forwarding based on overlay networks is enabled through the lightweight Segment Routing technology to provide higher performance on the forwarding plane. The network device layer simplifies the WAN architecture and protocol design using the SRv6 technology, and fully supports IPv6 network construction.

* Control layer: Through big data network analysis, SeerAnalyzer achieves network depth analysis and visualized presentation, network traffic prediction, security warning, fault location and other capabilities. SDN controller achieves centralized network resource tuning and scheduling, and centralized control. The control layer supports multi-network application integration to achieve cross-scenario unified scheduling, meeting the needs of different industry users. The control layer interconnects southbound with standard protocols and devices, and northbound provide standard user-centric APIs to achieve integration with third-party scheduling systems.

* Management arrangement layer: By calling the APIs provided by the orchestration layer app, it realizes policy definition, service issuance and management arrangement of services, such as real-time network monitoring, visualized presentation and troubleshooting. Thus, the visualized presentation of services and networks is optimized and the network O&M management is streamlined.

Features

Automatic service issuance

In the cloud computing era, new service models are merging, and enterprises have become digital in an all-around way. A notable feature of the cloud computing era is that users require services to have end-to-end fast online capabilities after services are migrated to the cloud. AD-WAN solution provides three kinds of abilities to automatically issue network services.

* VPN service provisioning: One-click L3VPN service is automatically provisioned. Users only need to specify the service deployment scope on the operation interface of the controller, and the configuration of controller the service to the network element device is automatically distributed by the controller, reducing the workload of manual configuration and the risk of operation and maintenance.

* QoS service deployment: End-to-end QoS services are automatically deployed to limit the traffic rate and ensure applications based on user services. The application-based end-to-end QoS guarantee function solves the problem of insufficient guarantee capability based on single-device deployment of QoS in traditional networks, and realizes rapid and batch delivery of service QoS configurations. When deploying QoS, users can configure one-click network-wide deployment based on a global template, or select a local template to deploy configuration for a single link, making the implementation more flexible.

* Service policy issuance: H3C AD-WAN industry solution performs the service policy issuance, such as bandwidth, quality, time period, and bandwidth calendar, automatically through controllers. And dynamic adjustment can be provided according to the real-time network status and service policy requirements to ensure reliable and stable service operation.

Flexible network scheduling

AD-WAN solution provides network scheduling options to achieve network optimization.

* Network scheduling based on the scheduling factors such as bandwidth utilization and link staining,

Assuming that two services go on the same link under normal circumstances, as the link bandwidth utilization gradually increases, the abnormal link service is automatically scheduled to another high-quality link based on next corresponding scheduling policy issued by the controller to guarantee user experience when the link utilization reaches the threshold.

* Network scheduling based on delay, jitter, packet loss rate and other network quality factors

Assuming that two services are on the same link under normal circumstances, the device automatically schedules the abnormal link service to another high-quality link based on the corresponding scheduling policy issued by the controller to ensure user experience when the delay of the link exceeds the threshold.

* Time period-based network scheduling

Based on the time policy, it can realize flexible network scheduling guarantee in different time sections, reasonably utilize network bandwidth resources, and ensure service reliability.

* UCMP multi-path loads

With the support of SRv6 Policy, the multi-path load balancing of service traffic can be realized. The weight can be specified by the administrator to define the load sharing ratio, and the controller can also automatically calculate the weight according to the remaining bandwidth resources of the path.


Multi-dimensional service guarantee

* Multi-dimensional strategy combination and application self-guarantee

Based on different strategies such as time period and bandwidth, different services are systematically combined to meet the needs in different user network scenarios. For example, based on the requirements of real-time applications such as video conferences, scheduling policies can be formulated based on comprehensive factors such as enabling delay, packet loss, and bandwidth utilization policies to perform flexible traffic scheduling and ensure the ultimate application experience.

* Network routing, application network solid state power amplifier (SSPA)

Users can statically select the exact path for applications to forward each hop between network nodes according to actual needs, and implement strict application routing and strict management and control. Optional or excluded nodes or links can also be specified for applications to implement loose routing of applications. The path can be pre-calculated before the scheduling policy is issued, which provides administrators with a reference of the application network SSPA. After confirmation, it is sent to the network to ensure the accuracy of the user's intention and reduce the risk of operation and maintenance.

* Network resource reconstruction, application optimization with one click

It can provide two key network capabilities based on single application path optimization and network-wide service optimization to avoid network congestion and forwarding deadlock in extreme cases, and improve network-wide resource utilization and risk resistance capabilities.

Smart network O&M

Smart SeerAnalyzer is introduced, and combined with the Telemetry technology, the rapid network status awareness and second-level O&M are fulfilled. Collect network-wide information (including network equipment, traffic, quality, related events and alarm information, etc.), conduct data mining, show users the most critical value elements in the network, assist users in IT decision-making, and assist in O&M. AD-WAN solution provides the following four key O&M capabilities:

* Network-wide service traffic visibility

It provides information views such as service bandwidth, flow direction, and health status in real-time, and displays it directly to end users, allowing users to always control the overall health status of the service.

* Overall device resource monitoring

It provides multi-dimensional topology display, allowing administrators to view network topology information from different perspectives and present an overall topology.

* Network health assessment

Based on the device dimension, evaluate network health, collect device operating status and service carrying status through Telemetry technology, and evaluate the current network operating status and health.

Based on link status assessment health, use telemetry technology to collect information such as link service carrying status, optical power throughput, etc., and establish a health baseline to assess network health.

* Application refined visibility

Based on the Netstream technology, the service traffic is visible, and the network traffic is collected to display the session-level traffic visualization.

Based on time, forward path visibility and historical path backtracking, support netstream-based service forwarding path visibility, and can backtrack historical paths.

* Intelligent analysis with trend prediction and simulation deduction

By analyzing the performance data such as devices, interfaces and optical links collected by the Telemetry characteristics of the network device, SeerAnalyzer can actively monitor and predict network anomalies and output prediction results, helping users to see the trend of network status changes, avoid faults, and prevent problems before they occur.

Link traffic trend forecast: You can view the predicted link bandwidth ratio by dragging the timeline. A traffic trend model is established based on ARIMA and LSTM model methods, which can accurately predict the changes of network traffic trends in the next 24 hours and 60 days, assist administrators in establishing intelligent traffic baselines, and refine network resource management.

Optical module fault prediction: Through the training and learning of the parameter changes of the optical module, based on the big data machine learning algorithm (logistic regression model), after the first-level alarm of the optical module is found, the future fault time of the optical module is predicted, and the fault warning point is given.

Simulation analysis of traffic and faults: SeerAnalyzer synchronizes the network data of the existing network to form a virtual "real" network. Based on this network, link and device fault simulation are performed to provide a basis for users to optimize network resource allocation in a timely manner.

Typical networking

Backbone network

Networking description:

* Network structure: Full-Mesh network, divided into different planes, and the planes are mutually backup.

* Application features: Vertical traffic and horizontal traffic, through VPN, priority, etc. to achieve business distinction and isolation.

Key needs:

* Optimize the backbone network traffic distribution and improve the link utilization.

* Improve the network reliability and guarantee the service quality of key services.

* Simplify the network O&M cost and reduce cost.


Vertical network

Typical networking:

Networking description:

* Network structure: Tree-shaped multi-level vertical network; the nodes are in dual-device dual-link redundancy, which can realize cross-domain connection with the upper-level network.

* Application features: Vertical application traffic is the main service, and the service differentiation and security isolation are realized through IP or VPN.

Key needs:

* Automatic deployment of equipment and services, automatic network service distribution, VPN, QoS and other configuration provisioning.

* Service traffic routing control based on flexible policies (such as time period, bandwidth, latency, jitter, packet loss rate, etc.) improves network reliability and ensures service quality of key services.

* In-depth network visibility (service, network, traffic, and path visualization), intelligent O&M, and simplified network management.

新华三官网