SRv6—A killer for 5G technology implementation

The development of 5G services has put forward higher requirements for network connections, such as stronger SLA, deterministic latency, etc. In the application scenarios of 5G, the business characteristics of various vertical industries are different. For wide-connection scenarios, such as smart home, environmental monitoring, smart agriculture, smart meter reading and other services, the network needs to support massive device connections and a large number of small message frequent services; for services such as video surveillance and online medical care, the network needs to support large-bandwidth and low-latency services; for services such as Internet of Vehicles, smart grids and industrial control, the network is required to support millisecond-level latency and reliability of no less than 6 9s. 5G technology applied to various industries must have greater flexibility and scalability.

5G service requirements can be divided into three categories:

• eMBB (Enhanced Mobile Broadband) focuses on services that have high bandwidth requirements, such as high-definition video and virtual reality/augmented reality;
• uRLLC (Ultra-Reliable Low-Latency Communication) focuses on services that are extremely sensitive to latency and reliability, such as autonomous driving, industrial control, telemedicine, and drone control;
• mMTC (Massive Machine Type Communication) focuses on covering scenarios with high connection density, such as smart cities and smart agriculture.

Figure 1: 5G service requirements

These three types of business scenarios put forward different network characteristics and performance requirements for 5G networks. Taking the bearer network as an example, the networking function of the 5G bearer network requires the realization of six requirements, including: multi-level bearer network, flexible connection scheduling, hierarchical network slicing, intelligent collaborative management and control, 4G/5G hybrid bearer and low-cost high-speed network. In order to meet the differentiated needs of different businesses in a physical network, network slicing technology must be used to achieve this. Network slicing meets different business needs by dividing a physical network into multiple virtual networks containing specific network functions, network topology and network resources.

SRv6 is an important part of the 5G technology system. Generally speaking, SRv6 has innovative applications in three aspects of 5G scenarios: network slicing, service management, and cloud resource management.

Application of SRv6 in Network Slicing

5G network slicing involves terminals, wireless, bearer and core networks, and requires end-to-end collaborative management and control. Through resource slicing on the forwarding plane and slicing management and control on the control plane, differentiated SLA guarantees are provided for the three major types of services.

Slices need to have the following characteristics:

• Meet different business needs: Provide differentiated services on demand.
• Security: Effectively isolate data/information between tenants.
• Reliability: Network anomalies of any tenant will not affect other tenants in the same physical network.

Network slicing is the specific manifestation of network function virtualization applied to the 5G stage. A network slice constitutes an end-to-end logical network, which flexibly provides one or more network services according to the needs of the slice demander. The network slicing architecture mainly includes two functions: slice management and slice selection. The slice management function organically combines operations, platforms, and network management to provide secure isolation and highly controllable dedicated logical networks for different slice demanders. The slice selection function realizes access mapping between user terminals and network slices. The slice selection function integrates multiple factors such as service contracting and functional characteristics to provide user terminals with appropriate slice access options.

Taking the bearer network technology SPN as an example, the SPN network slicing layered architecture includes three levels: slice packet layer (SPL), slice channel layer (SCL) and slice transport layer (STL).

Figure 2: SPN protocol stack

SPN uses the slice packet layer to carry CBR services, L2 virtual private networks, and L3 virtual private networks. The slice packet layer is centrally managed based on SDN and provides connection-oriented and connectionless communication carrying capabilities. SRv6 is an important implementation technology of the slice packet layer.

SRv6 Policy uses the SR mechanism to encapsulate an ordered list of instructions at the head node to guide the message to traverse the network. SRv6 uses Color to identify the SRv6 Policy ID. This parameter is associated with business attributes, such as low latency, high bandwidth, and other business attributes. This parameter can be understood as the template ID of the business demand. There is currently no unified encoding rule for the Color identifier, and the value is assigned by the administrator. For example, a policy with an end-to-end delay of less than 5ms can be assigned a Color identifier value of 50. Through this parameter, SRv6 Policy can directly respond to business needs, thereby omitting the process from business needs to network language and then to network objects.

SRv6 is associated with the specified network slice through the Slice ID information in the HBH (Hop-by-Hop Options Header) extension header. When the slice data reaches the SRv6 routing source, it is associated with the SRv6 TE Policy according to the VPN instance routing table, and then the SRH information is inserted into the message, the SID List of the SRv6 TE Policy is encapsulated, and then the HBH extension header is encapsulated, and finally the basic IPv6 message header is encapsulated. The slice data is encapsulated into a standard IPv6 message and then forwarded. When forwarding, it is associated with the specified network slice interface through the Slice ID information.

Figure 3: SRv6 slice encapsulation

An SRv6 Policy may be associated with multiple candidate paths, supporting primary and backup redundant paths and escape paths. SRv6 Policy supports multiple path calculation methods, including PCE centralized path calculation, head node path calculation, manual path planning, and FlexAlgo path calculation. The design of multiple candidate paths shields the service from path calculation details. Paths formed by different path calculation methods are encapsulated in the SRv6 Plolicy in the form of candidate paths.

Currently, the SRv6 SID list uses 16-byte IP address identifiers, which has low forwarding efficiency. SRv6 has high requirements for forwarding chips. Currently, only a small number of chips can achieve 10+ layers of label encapsulation. In addition, there are still devices that do not support SRv6 in the carrier-class bearer network. MPLS and SRv6 will coexist for a certain period of time. It is necessary to achieve coexistence between different technologies through MPLS and SRv6 splicing solutions, dual-plane solutions, and overlay solutions.

In 2020, ZTE supported China Telecom in realizing the world's largest SRv6 commercial network and achieved 2B/2C slicing deployment, introduced MPLS and SRv6 splicing and dual-plane technology, and realized IP RAN1.0 and STN integrated commercial deployment.

SRv6 support for 5G service management

The multi-service carrying of 5G networks has put forward higher requirements for network management. More effective fault location means are needed to improve operation and maintenance efficiency. The granularity of traditional performance monitoring OAM is relatively coarse, and network quality is often indirectly detected by constructing monitoring messages. This method can only monitor ports/tunnels/pseudo-wires, and the information collection cycle is at the minute level, which cannot capture sudden anomalies. In addition, traditional monitoring methods cannot automatically monitor hop by hop and cannot quickly demarcate faults. Traditional performance monitoring cannot meet the needs of 5G service management. In 5G scenarios, a performance monitoring method based on the service flow level needs to be provided. 5G monitoring methods need to have a real-time, high-precision performance detection mechanism. At the same time, for delay-sensitive services, performance monitoring also needs to provide end-to-end delay conditions for the entire network, be able to monitor delay anomalies, and be able to adjust the routing for abnormal paths in a timely manner.

SRv6-based network measurement instu-OAM has become an IETF working group draft. SRv6-based instu-OAM can provide millisecond-level network node service perception capabilities, adapt to the development of network intelligent routing, and ensure telecom-level service reliability.

The on-link network measurement based on the IFIT (In-situ Flow Information Telemetry) framework does not introduce additional measurement messages, can monitor user flow delay and packet loss in real time, can quickly find fault points, supports data plane encapsulation of multiple on-link network measurement technologies, and can be deployed on a large scale in IP networks. On-link network measurement does not actively send detection messages, and measures network performance by carrying OAM instructions in user messages. The on-link network measurement method measures real user traffic, can achieve detailed monitoring of messages, and can collect the forwarding path of messages in the network and the hit rules of each hop device.

IOAM supports two modes: Passport and Postcard. The Trace Option defined by IOAM implements the routing of Passport mode. The Trace Option includes 64 bits.

Figure 4: IOAM Trace Option header

The IOAM-Trace-Type 24 bits are used to describe the data that needs to be collected, and each bit represents a type of data that needs to be collected. Different manufacturers can define the supported data types under the specified namespace. These data include node identifiers, receiving message interface identifiers, sending message interface identifiers, message processing time in the device, unformatted opaque data, etc. Every time a message passes through a node in an IOAM domain, the node will collect the corresponding type of data according to the IOAM-Trace-Type and add it after the IOAM instruction once.

EAM provides postcard mode on-line network measurement. EAM instructions have 32 bits, of which FlowMonID has 20 bits, which is used to identify the monitoring flow. This field is written by the ingress node of the monitoring domain.

Figure 5: IOAM EAM Header

IOAM implements PBT-I (Postcard-Based Telemetry with Instruction Header) by adding the IOAMDEX (Directly EXport) option. In this mode, the instruction header is 128 bits long, the same as the IPv6 address length, and can be carried in the IPv6 extension header.

With the help of SRv6 technology's programmability in the data plane, IFIT instructions can be encapsulated in the IPv6 HBH or in the IPv6 SRH extension header. The former IFIT instructions will be processed by all IPv6 forwarding nodes. The latter will only be processed by the specified Endpoint node.

Figure 6: Two ways of encapsulating management fields

The management and control system needs to have the planning and optimization functions of multi-layer network resources, achieve the optimal configuration of multi-layer network resources, and realize the coordination of multi-layer routing strategies.

Application of SRv6 in 5G cloud resources

The development of cloud technology has made the location of business processing more flexible. Cloud services break the boundaries between physical network devices and virtual network devices, and services and bearers are integrated.

5G edge cloud integrates DCN and WAN network to form Spine-Leaf Fabric architecture. Leaf is the access node of Fabric network function, usually the PE device in WAN network. Spine mainly forwards high-speed traffic, usually the P device in WAN network. Leaf nodes are connected through high-speed interfaces, and a larger network is covered through hierarchical interconnection.

Figure 7: 5G cloud network architecture

Edge cloud connects DCN and WAN, and the transmission bearer protocol also needs to be connected. SRv6 is an effective way to solve the problem of connecting DCN and WAN networks. The SRv6+E virtual private network technology realizes IPv4/v6 dual-stack service capabilities and provides service capability support for 5G, enterprises and MEC. Through end-to-end SRv6 BE/TE, overall path optimization is performed and service isolation is achieved. SRv6 eliminates the service configuration point of Option A of the cross-domain virtual private network between back-to-back DC-GW and PE, and provides end-to-end OAM capabilities. At the same time, SRv6 makes it possible to simplify the network hierarchy. There is no need to independently set up multi-layer nodes such as PE, DC-GW, Leaf, etc. between DCs. These functional devices can be merged together and multiplexed by a layer of devices, reducing the cost of network construction.

In addition, in terms of SRv6 forwarding efficiency, China Mobile proposed the "G-SRv6 header compression optimization solution", which achieves compression by removing redundant prefixes in SRv6 SIDs, effectively solving problems such as excessive native SRv6 header overhead, low forwarding efficiency, and high requirements for header processing hardware, thus removing the biggest obstacle to large-scale deployment of SRv6.

summary

As 5G is more widely used in more scenarios, SRv6 will play an increasingly important role. L3 network slicing, on-path service monitoring, and cloud-network integration based on SRv6 will also become an important cornerstone of 5G network construction.