Segment Routing

What Is Segment Routing?

Segment Routing (SR) is a source routing protocol, fundamentally designed to optimize packet forwarding processes. Its key concept involves segmenting a packet's forwarding path into discrete segments. Segment information is then embedded into packets at the entry point (ingress) of the path, providing instructions for subsequent packet forwarding actions. SR accommodates both MPLS and IPv6 data planes, representing two prevalent technical approaches: SR-MPLS and SRv6. SR-MPLS refers to SR implementation on the MPLS data plane, while SRv6 denotes SR deployment on the IPv6 data plane.

How Does SR Come into Being?

The rapid advancement of global informatization is propelling the growth of Internet applications. With the proliferation of networks and the advent of the cloud era, there is a burgeoning diversity in the spectrum of network services and their associated requirements. Against this backdrop, traditional IP/MPLS networks encounter a myriad of challenges:

• Fragmented IP bearer network islands: Despite the unification of bearer network technologies by MPLS, IP backbone, metro, and mobile bearer networks exist as separate MPLS domains, resulting in isolated islands. Consequently, intricate technologies like inter-AS VPN are required to interconnect them, leading to complex end-to-end service deployment. Moreover, the coexistence of L2VPN and L3VPN services necessitates multiple protocols (e.g., LDP, RSVP, IGP, and BGP) on a single device, further complicating management and hindering large-scale service deployment.

• Limited programming space in IPv4 and MPLS: Modern services often require additional forwarding information to be included in packets. However, the IETF's decision to cease formulating new standards for IPv4 exacerbates the challenge. Furthermore, the fixed 20-bit length of the MPLS label space lacks extensibility, rendering it inadequate for meeting the programming needs of emerging services.

• Disconnection between applications and bearer networks: The disconnection poses challenges in optimizing networks and deriving maximum value from them. Consequently, many carriers find themselves limited to providing basic connectivity services, unable to leverage value-added applications. Additionally, the absence of application insight restricts carriers to coarse-grained network adjustment and optimization, resulting in resource inefficiencies. Despite various attempts over the years to integrate MPLS closer to user hosts and applications, such efforts have faltered due to factors like numerous network borders and intricate management processes.

• Tight integration of the data and control planes: The entwining of these planes for sales and evolution purposes prolongs service provisioning timelines and complicates adaptation to the rapid evolution of new services.

In response to these challenges, several innovative network technologies have emerged, with Software-Defined Networking (SDN) standing out prominently. Traditional network architectures are inherently characterized by tightly integrated hardware devices, operating systems (OSs), and network applications, leading to various limitations. To address these issues, Professor Nick McKeown's team at Stanford University proposed SDN, a new network architecture leveraging universal hardware, software-defined functionalities, and an open-source model from the computing field. SDN is distinguished by three key characteristics: network openness and programmability, logical centralized control, and separation between control and forwarding planes. Any network exhibiting these traits can be categorized as an SDN network.

Segment Routing (SR) was developed under the influence of the SDN concept. Its fundamental principle involves segmenting a packet's forwarding path and embedding segment information into packets at the path's ingress. Transit nodes then only need to forward packets based on the carried segment information. As a straightforward, efficient, and scalable protocol, SR offers several advantages, including:

• Facilitates network path programming: SR leverages the inherent benefits of source routing, allowing the ingress to govern a service path simply by manipulating packet labels. Moreover, transit nodes are relieved from the burden of maintaining path information, thereby alleviating strain on the control plane.

• Streamlines device control plane: By reducing the need for multiple routing protocols and simplifying the label forwarding table, SR minimizes device resource utilization and operational costs.

• Seamless transition to SDN networks: With a design aligned with SDN principles, SR combines the advantages of autonomous forwarding and centralized programming control to facilitate application-centric networks. Additionally, it supports both traditional and SDN environments and is compatible with existing infrastructure, ensuring a smooth transition to SDN networks.

What Technical Solutions Does SR Support?

The operational principles of SR can be likened to real-life scenarios. Consider planning a journey from Shanghai to Paris with a layover in Vienna. Your travel route is segmented into two parts: Shanghai to Vienna, and Vienna to Paris. To reach your final destination, you only need to purchase a ticket to Shanghai. By taking two flights, you'll arrive in Paris as intended.

Similarly, in SR, the process of forwarding network packets follows a comparable pattern. It involves segmenting a path and organizing these segments (segment list) at the initial point to establish the forwarding route. SR amalgamates segments representing distinct functions into a list, enabling path programming and catering to diverse service quality requirements.

SR accommodates both MPLS and IPv6 data planes, corresponding to SR-MPLS and SRv6, two predominant technical solutions. SR-MPLS operates on the MPLS data plane, while SRv6 operates on the IPv6 data plane. For further insights into these technical solutions, refer to SR-MPLS and SRv6 documentation.