In the previous post, we introduced the problems and challenges faced by WAN backbone networks and described an ideal architecture for future WAN backbone networks. In this post, we will describe how the SDN-based IP+optical solution narrows gaps between dream and reality.

Through years of research and its deep understanding of IP and optical technologies, Huawei has developed a unique SDN-based IP+optical solution. This solution aims to convert the WAN backbone network to a software-defined network that can meet various service requirements.

The WAN backbone network uses the IP bearer technology to forward IP packets and uses the optical bearer technology to transmit optical signals through optical fibers over long distances.

The two technologies complement each other and display their respective advantages in a unified management and control system. In addition to improving traffic visibility, the SDN-based IP+optical solution enables WAN backbone networks to be more agile, efficient, integrated, and open, thereby improving user experience.

SDN-based IP+optical solution

 

The preceding figure shows an IP+optical network that:

• Improves resource usage and traffic scheduling efficiency by using IP and optical path computation elements (PCEs) to compute cross-layer paths and manage policies.
• Improves network deployment efficiency by converging the control planes of IP and optical networks through Generalized Multiprotocol Label Switching (GMPLS) user-to-network interface (UNI).
• Implements centralized service configuration and provisioning through the NetMatrix.
• Enables visualized display of traffic distribution by using the uTraffic to uniformly monitor network traffic.
• Improves network planning efficiency with a multi-layer planning tool.

The key to this solution is integration between the control and orchestration layers. Because the use of centralized software to define resources and services on the entire backbone network matches the core concept of SDN, the architecture used by this solution is also named SDN-based IP & Optical Synergy.

Still confused? Lets have a close look at this solution.

Solution Description

Optical Network: Building a Flexible Bandwidth Pool

To achieve agility, high efficiency, and good user experience, the WAN backbone network must have a flexible bandwidth pool to rapidly provide bandwidth resources based on service requirements. In addition, the WAN backbone network must have an automated implementation mechanism to allocate physical resources. MS-OTN devices are optical network devices that meet various technical requirements of such a bandwidth pool.

MS-OTN devices can flexibly schedule services carried over wavelengths at Layer 0, over ODUk channels at Layer 1, and over Multiprotocol Label Switching (MPLS) tunnels at Layer 2.

At Layer 0, MS-OTN devices use wavelengths to transmit coarse-grained services over long distances. Despite the bandwidth capability of a single wavelength constantly improving, requirements for end-to-end bandwidth pipelines vary according to services. Subsequently, mapping low-speed pipelines to high-speed wavelengths becomes a daunting challenge to the optical network.

At Layer 1, MS-OTN devices use optical transport network (OTN) encapsulation and centralized cross-connect technologies to multiplex and schedule low-speed signals onto high-speed wavelengths, improving wavelength bearer efficiency. ODUflex improves the traditional ODUk (k=0, 1, 2, 3, 4) inter-layer mapping mode and provides matching timeslots for constant bit rate (CBR) and variable bit rate (VBR) services. ODUflex further prevents resource wastage and improves wavelength bearer efficiency. With a granularity of 1.25 Gbit/s, ODUflex can dynamically update circuit bandwidth by adding or deleting timeslots in ODUflex channels.

In addition to Layer 0 and Layer 1 functions, MS-OTN devices can use Multiprotocol Label Switching Transport Profile (MPLS-TP) to bear and schedule Layer 2 packet services. By identifying VLANs and mapping VLAN packets to ODUflex channels, ODUflex allows circuits with desired bandwidth to be constructed between MS-OTN devices and routers on demand, preventing resource wastage.

MS-OTN devices also support automatically switched optical network (ASON)-based automated resource scheduling. At the core of ASON is the GMPLS control plane, which is used by MS-OTN devices. This control plane is like a powerful brain for the MS-OTN devices. It streamlines network management on the network management system (NMS) and improves network reliability, as well as enabling the optical network to automatically discover resources and rectify faults. ASON uses restoration and rerouting mechanisms to protect optical services against unexpected failures.

IP Network: Flexibly Interoperating with the Optical Network

Backbone edge routers receive a variety of services to be scheduled by the WAN backbone network. These routers generate IP routing tables to guide traffic forwarding along specific IP routes. To forward IP traffic, routers must have IP links between each other.

Currently, most IP links are established between physical router interfaces. As the transmission rates of these interfaces connecting to optical network device interfaces are constantly increasing, physical-interface-based IP links face the problem of rigid and coarse-grained traffic scheduling.

To address this problem, configure VLAN sub-interfaces on each high-speed physical router interface and establish IP links between VLAN sub-interfaces. The logical-interface-based IP links will then distribute traffic based on VLANs to flexibly schedule traffic, improving bandwidth usage.

Because MS-OTN devices support Layer 2 features, they can identify logical sub-interfaces based on identifiers such as VLAN IDs and transmit IP traffic to each destination along specific ODUflex channels. In addition, logical sub-interfaces and ODUflex channels both support flexible bandwidth settings, which enables logical-interface-based IP links to dynamically adjust bandwidth according to traffic changes.

Logical-interface-based IP links can also aggregate traffic from multiple remote physical interfaces to a local physical interface or bypass traffic without requiring additional physical interfaces. The VLAN+MS-OTN architecture used by the SDN-based IP+optical solution allows for flexible end-to-end channels on backbone networks.

To meet IP link connection and bandwidth adjustment requirements, routers must be able to instruct optical network devices to provide matching resources. The routers and optical network devices use GMPLS UNI for communication. Before a router establishes, tears down, or modifies an IP link connecting to an optical network device, the router sends GMPLS UNI signaling to the optical network device. Upon receipt of the information, the optical network device automatically sets cross-connect and optical parameters.

SDN Controller: Performing Centralized Control from a Global Perspective

The key to IP+optical convergence is centralized management of the IP and optical networks for unified resource scheduling and end-to-end service deployment.

IP routing protocols used on the control plane of the IP network generate IP routes or MPLS forwarding labels based on the activated IP link topology.

GMPLS used at the control plane of the optical network automatically completes end-to-end optical-layer path computation, cross-connect of each node on the path, and configuration of optical parameters for each node based on the service requirements.

GMPLS UNI can transmit IP link connection requirements to the optical network; however, it requires manual configuration of IP link connection parameters. IP/MPLS and GMPLS both use distributed route management. To be specific, a device maintains only its own forwarding table and information about services with the device as the source or sink. Distributed information management does not support global load balancing or resource usage optimization. To implement global load balancing and resource usage optimization, each device residing on a backbone network with thousands of devices has to have high hardware performance to store network-wide link resource and service information and compute cross-layer service paths.

The SDN-based IP+optical solution uses an SDN controller to globally process data and compute paths. This implementation facilitates cross-layer optimization and control.

The IP and optical PCEs are integrated into the IP+optical controller.

The IP controller performs functions such as collecting IP layer network resource information, computing paths for TE tunnels, initiating path computation requests to the optical layer, managing Virtual Network Topology (VNT), and implementing path computation policies. The optical PCE computes paths for UNI tunnels at the optical layer based on the optical layer topology, resources, and configurations.

Unified O&M Plane: Implementing Network Abstraction and Automated Service Deployment

On the backbone network, IP traffic keeps changing. The changes in IP traffic paths, bandwidth, and directions must be controlled in real time for subsequent resource adjustment. To meet this requirement, the SDN-based IP+optical solution uses uTraffic to collect and analyze backbone network traffic information, improving network traffic and performance visibility.

To implement backbone network integration, the SDN-based IP+optical solution uses the NetMatrix for unified resource, service, and policy management and service provisioning.

The NetMatrix supports such functions as global resource and topology management, automated end-to-end service provisioning, unified policy management, and global VNT and shared risk link group (SRLG) management. This platform not only integrates resources, but also abstracts implementation details to provide user-friendly graphical user interfaces (GUIs).

The uTraffic and NetMatrix provide an important platform for network resource abstraction, traffic visualization, and service provisioning automation.

The NetMatrix supports the following functions:

• Network and service deployment
• Policy-driven online optimization
• Open APIs for programmability

The NetMatrix consists of the following modules:

• Policy-Engine: triggers network optimization based on the real-time network status and user-defined policies.
• ML-Optimization: implements cross-layer optimization and local service adjustment.
• ML-Provision: implements inter-layer and end-to-end network and service deployment.
• Modeling-based Development (MBD): dynamically generates RESTful open APIs for NE and drive adaptation (including third-party adaptation) by adding plug-ins.
• VNT-Management: manages the VNT, such as importing VNT planning and dynamically establishing and deleting VNT links.
• SRLG-Management: manages SRLGs, such as importing user-defined SRLG configurations.

The uTraffic supports the following functions:

• Traffic information collection, analysis, and forecast
• Visual display of traffic and network performance information
• Traffic matrix shaping
• Elephant flow identification

The uTraffic consists of the following modules:

• Collector: collects traffic/performance statistics.
• Analytic: analyzes traffic information.
• Forecast: forecasts the traffic trend.

Multi-layer Planning Tool: Implementing Cross-layer Planning

Network planning, deployment, and emulation require planning tools. Nowadays, the market is flooded with planning tools that can help carriers plan their networks. However, these planning tools are offline tools and apply only to planning based on traffic forecast results. In addition, these planning tools are intended for a single domain and do not consider the IP and optical networks as a single entity.

In the SDN architecture, the uTraffic and NetMatrix work together to collect and aggregate network resource and traffic information in real time. Therefore, it is possible to implement online planning based on real-time network resource status and traffic information. The multi-layer planning tool can be regarded as an NBI-based value-added application. It computes network optimization policies and forecasts network failures based on the network topology, service, and traffic information obtained in real time, providing suggestions for network planning and maintenance.

The multi-layer planning tool supports topology, capacity, service path, resource, and policy planning.

The inputs for this multi-layer planning tool can be:

• Live network resource information
• Live network matrix information
• Service requirements
• SLA requirements

Information that can be imported to the multi-layer planning tool includes network resource information from the NetMatrix and traffic matrix information from the uTraffic. The multi-layer planning tool then automatically implements IP and optical layer network modeling and establishes the network topology and model.

Multi-layer planning performed after the network model is established includes:

• IP layer planning: includes IP topology and TE planning.
• IP-optical mapping: includes the planning of optical layer resources, optical topology, VNT, collaborative protection, SRLGs, and policy constraints.
• Traffic forecast and traffic/capacity analysis: is implemented based on current and potential service requirements to determine link usage and network bottlenecks.

After cross-layer planning, the multi-layer planning tool simulates faults to verify the robustness of affected traffic. The fault simulation function can simulate either a single point of failure, such as a fiber, device, or SRLG fault, or a compound fault, such as simultaneous device and link faults.

Multi-layer planning is an iterative process that gradually approaches an ideal status. The tool implements automated planning based on predefined service requirements and constraints. If the requirements and constraints cannot both be met (for example, the target costs and SLA requirements conflict), the user must adjust requirements or constraints based on the current situation to meet the planning target.

The tool obtains abstract network information, such as port quantity and cross-connect capacity, from the NetMatrix and uTraffic. Because the abstracted information is independent of vendors, the tool is suitable for multi-vendor scenarios.

Solution Usage Scenarios

Currently, this solution can be applied to the following scenarios:

• Automated multi-layer network deployment
• Automated service provisioning
• Multi-layer protection
• SRLG-based path disjoint

This solution enables carriers to construct an open, intelligent, agile, and flexible WAN backbone network that features centralized cross-layer management and automated cross-layer service deployment, helping carriers reduce total cost of ownership (TCO) and accelerating service innovation.

In subsequent posts, we will describe each usage scenario one by one. 。

This post describes only the general design idea of the SDN-based IP+optical solution. For specific functions and implementations, see the user documents.
The multi-layer planning tool and functions related with uTraffic will be support later.

Terms

OTN: optical transport network. The OTN is a WDM-based next generation backbone transport network.

MS-OTN: a new generation OTN product released after NGWDM products with support for MPLS-TP. These devices are OTN devices that support MPLS-TP and packet switching.