Role of Time-Sensitive Networking (TSN) for Fronthaul in 5G Networks

Learn how time-sensitive networking (TSN) for fronthaul can help you meet the transport convergence, capacity, latency, synchronization, and resiliency requirements of 5G networks.

Role of Time-Sensitive Networking (TSN) for Fronthaul in 5G Networks
Role of Time-Sensitive Networking (TSN) for Fronthaul in 5G Networks

Learn how Time-Sensitive Networking (TSN) for fronthaul can:

  • Address 5G service diversity and time-sensitive applications
  • Deliver bandwidth-efficient, packetized transport for C-RANs
  • Converge transport for all 4G/5G traffic types
  • Provide highly precise network-based synchronization
  • Enable ultra-low latency transport with deterministic forwarding
  • Support carrier-grade OA&M, availability and resiliency

Content Summary

Introduction
5G Service Diversity
5G Ran Architecture Evolution and Transport Requirements
Functional Splits and Fronthaul
5G and Edge Cloud
TSN Fronthaul to the Rescue
Converged Networks
High Precision Synchronization
Ultra-Low Latency
Carrier OA&M and High Resiliency/Availability
Future and Industry Next Steps

Introduction

The early drumbeat of global 5G commercial announcements is a sign of encouraging progress, but these announcements do not convey the full picture. The true promise of 5G can be realized only after standalone 5G is standardized in Release 16, and the changes will be significant. Meeting the requirements of standalone 5G with networks that scale for mass-market adoption is a massive undertaking that is just now beginning. Beyond the radio, transport networks will play a crucial role in this 5G network evolution, and a host of new technologies are coming to market to meet 5G’s unique demands.

The IEEE’s 802.1CM Time-Sensitive Networking (TSN) for Fronthaul standard and its upcoming 802.1CMde amendment is among the most promising new Ethernet transport technologies developed especially for 5G networks. This white paper provides an overview of TSN for Fronthaul with an emphasis on what is driving the need for a packetized fronthaul transport option and how Ethernet meets fronthaul requirements in convergence, capacity, ultra-low latency, synchronization, operations, and resiliency.

5G Service Diversity

The diversity of 5G’s use cases and deployment options sets 5G apart from previous mobile technology generations. The 3GPP has defined three broad categories of use cases that 5G must support: enhanced mobile broadband (eMBB), massive machine-type communications (mMTC), and ultra-reliable low latency communications (URLLC).

Characteristics and performance requirements vary widely across the use cases. eMBB bears the most similarities to today’s 4G, but capacities increase by an order of magnitude or greater. Indoor stadium broadband access may require 1 Gbps download rates to users, as one example. For mMTC (which encompasses the Internet of Things), data rates may be quite low, but the ultra-high density and ultra-low energy consumption become paramount. One million device connections per square kilometer may be required for some applications. Finally, in URLLC, six-nines levels of reliability and millisecond latencies may be required (see Figure 1).

Figure 1: Diversity of 5G Use Cases and Requirements. Source: Qualcomm
Figure 1: Diversity of 5G Use Cases and Requirements. Source: Qualcomm

Each broad use case category is populated with many 5G applications – some well-defined, others still largely conceptual, and many more yet to be conceived. Reality is more complicated than taxonomy diagrams depict because many of the 5G services will not fit neatly into a single-use case category. A smart factory, for example, may need ultra-low latency communications for its production lines but also use delay-tolerant sensors to keep track of inventory.

5G Ran Architecture Evolution and Transport Requirements

As 5G services and applications will not be one-size-fits-all, the same is true with the radio technologies that deliver 5G services. Although not officially defined in the industry, Heavy Reading categorizes three radio spectrum options as low band (sub-2 GHz), mid-band (2 GHz-6 GHz), and high band (anything above 6 GHz) spectrum.

The low band spectrum provides the greatest geographic coverage but also delivers the lowest data capacity. In the low bands, capacity is scarce, and maximizing spectral efficiency will be paramount (i.e., operators need to squeeze as much data as possible from each megahertz of spectral bandwidth). The spectral efficiency requirement is driving some operators toward centralized RAN (C-RAN) architectures that provide tight coordination between macro and street cells. Due to the favorable propagation characteristics of the low band spectrum, techniques such as massive MIMO (mMIMO) and beamforming are not needed. In this case, 5G radios can use the Common Public Radio Interface (CPRI) protocol (with time-division multiplexing [TDM] scaling). CPRI does not require L1 PHY processing at the radio unit (RU), thus helping to minimize the size and power requirements of the radios.

Mid-band spectrum ranges are suitable for metro coverage areas, and channel sizes in the 100 MHz range allow operators to increase data rates beyond 4G. Operators can use carrier aggregation for even higher aggregate data rates or use mMIMO antenna arrays to boost single-channel throughputs and achieve cell sizes consistent with 4G technology.

The highest data rates will be delivered in a high band spectrum, with particular interest globally in the spectrum above 24 GHz, also known as the millimeter wave (mmWave) bands. At frequency bands over 10 GHz, the time-division duplexing (TDD) spectrum is preferred, as it delivers better performance in MIMO/beamforming algorithms compared to frequency-division duplex (FDD). TDD spectrum needs high precision phase/time synchronization, as discussed in the High Precision Synchronization section of this paper.

For bands in the mmWave range, channel sizes range from 50 MHz up to 400 MHz, thus providing maximum 5G data rates without carrier aggregation. But high band spectrum options will have limited coverage, which drives the need for 4G and 5G interworking. 4G can provide broader coverage and 5G can provide high capacity at hot spots. Densification will come primarily in the form of small cells deployed on light poles, on buildings, within buildings, and on new fixtures – all places that do not necessarily have the existing 4G infrastructure.

Finally, it is important to understand that 5G and 4G will coexist for many years. Early implementations of 5G will be non-standalone, meaning that these 5G networks will share the 4G core. Even for standalone 5G implementations, interworking with the installed 4G base is needed to provide the necessary coverage (as described for mmWave). This coexistence of 4G and 5G dictates that the transport network must support both mobile generations.

Functional Splits and Fronthaul

In 5G, base station (gNB) functionality is split into three functional units: the centralized unit (CU), the distributed unit (DU), and the RU, which can be deployed in multiple location combinations. By centralizing baseband unit (BBU) functions in a C-RAN architecture, operators can share network resources and tightly coordinate radioactivity to improve network performance. The split functions create new transport links; specifically, the fronthaul link connecting the RU to the DU and the midhaul link connecting the DU to the CU. The backhaul link, connecting the CU to the mobile core, serves the same function as in previous mobile generations.

While improving radio network performance, the functional splits place new requirements on the transport network and, in particular, on fronthaul. The evolved CPRI (eCPRI) protocol was developed by the CPRI Cooperation to enable fronthaul networks to scale with 5G demands. By keeping some physical layer processing in the RU (see Figure 2), eCPRI reduces the transport bandwidth connecting the RU to the DU by as much as 10x compared to legacy CPRI. As a pocket-friendly protocol, eCPRI can be supported by Ethernet-switched and IP-routed networks, provided these packet networks to meet the performance requirements detailed in the eCPRI specification. Although eCPRI greatly improves bandwidth efficiency compared to CPRI, it is subject to the same tight latency requirements as CPRI.

Figure 2: RAN Decomposition Highlighting Fronthaul Functional Split. Sources: Heavy Reading, 3GPP
Figure 2: RAN Decomposition Highlighting Fronthaul Functional Split. Sources: Heavy Reading, 3GPP

5G and Edge Cloud

Edge computing places high performance compute, storage, and network resources as close as possible to end-users and devices with the goals of reducing the costs of transport, decreasing latency, and increasing locality. Low latency is a cornerstone of the URLLC set of use cases and one of the primary differences between 5G and previous mobile generations. Transport networks have a role to play, as distance and latency are directly related and any network processing adds further delay along a route. But without moving to compute and storage resources closer to users, ultra-low latencies cannot be achieved. Emerging 5G applications in artificial intelligence, drones, virtual and augmented reality, automated manufacturing, smart cities, and many others all require edge computing at some level.

Network operators clearly understand the importance of edge computing for 5G. In 2019, Heavy Reading conducted a global survey of network operators on the topic of edge computing. In that survey, 82% of operators reported that edge computing is at least important to their business. And 22% of the respondents reported that edge computing is critical and that their business cannot succeed without it.

Some operators are building on the work done in C-RANs in which the CU often sits at the base of a tower rather than in the cloud. For example, U.S. operators already have thousands of C-RAN hubs throughout the country that may be valuable physical locations for edge computing deployments and may create relatively localized virtualized RANs (vRANs). These C-RAN locations can serve as a stepping stone to future vRANs.

TSN Fronthaul to the Rescue

TSN is part of the IEEE 802.1 family of standards and is designed to provide deterministic forwarding on standard Ethernet networks. It is a Layer 2 technology that can be centrally managed and use coordinated scheduling to ensure performance for real-time applications. Real-time deterministic communications are important to many industries – for example, aerospace, automotive, transportation, utilities, and manufacturing. In many of these industries, TSN is emerging as the baseline for real-time networking.

Published in 2018, the TSN Fronthaul standard (IEEE 802.1CM) profiles IEEE 802 standards and ITU-T synchronization recommendations specifically for fronthaul transport networking in a C-RAN architecture. The standard specifies an Ethernet bridged network for connecting radio equipment (RU functionality) to a remote controller (DU functionality). TSN for Fronthaul is specified by the IEEE 802.1 Working Group but was developed in collaboration with the CPRI Cooperation, which provided the CPRI and eCPRI requirements.

Synchronization work from the ITU-T’s Study Group 15 was also referenced in developing IEEE 802.1CM. The Study Group has defined the network-based synchronization profiles for packet transport networks, including Ethernet in the fronthaul.

Heavy Reading research shows that eCPRI, network synchronization, and TSN for Fronthaul are all critical technologies in operators’ 5G transport network plans. In a 2020 Heavy Reading global survey, we asked network operators to rank the most important technologies and protocols for their 5G transport network plans. In terms of importance, network synchronization protocols, eCPRI, and TSN for Fronthaul were three of the top four technologies selected (see Figure 3).

Figure 3: How important are the following technologies and protocols for your 5G transport network? (Ranked by percentage important or critical for each). Source: Heavy Reading 5G Network & Service Strategies: 2020 Operator Survey
Figure 3: How important are the following technologies and protocols for your 5G transport network? (Ranked by percentage important or critical for each). Source: Heavy Reading 5G Network & Service Strategies: 2020 Operator Survey

The hybrid automatic retransmit request (HARQ) protocol typically limits the fronthaul transmission delay between the RU and the baseband processing to 100 μs. With fiber-imposed latency at 5 μs/km, physics restricts fronthaul connectivity to 20 km or less. Because of these restrictions, fronthaul links to date have been point-to-point connections over dark fiber or wavelengths on fibers – regardless of whether the fronthaul protocol is CPRI or eCPRI. Intermediate processing adds latency, particularly in switched and routed networks.

The TSN for Fronthaul standard specifies two transport profiles, both of which apply to CPRI and eCPRI protocols. Profile A sends user data (IQ data) as a high priority class above control and management data. Profile B includes components of Profile A but also adds a TSN feature called frame preemption (originally specified in IEEE 802.3br and 802.1Qbu) to prioritize different traffic types. As discussed further in the Ultra-Low Latency section, Profile B adds a greater level of performance by better bounding latency.

Converged Networks

GSMA forecasts that 4G will still account for over half (56%) of the technology mix by 2025; 5G will account for only 20%, stressing the need for converged transport. The key advantage of IEEE 802.1CM is in building a converged mobile transport network. Connecting each radio with one or more fibers or more wavelengths can be wasteful, though at times the most economical option. With legacy CPRI (which is a serial protocol), there is no other option unless mapped using Radio over Ethernet (RoE), but eCPRI is a packetized protocol that allows for aggregation. Using TSN for Fronthaul, time-sensitive traffic can be transported in the same packet network as non-time-sensitive traffic without being negatively affected. This enables efficient traffic aggregation into high data rate ports (such as filling a 100 gigabit Ethernet link) and allows a single TSN node to serve multiple types of traffic.

Supported traffic types include traditional CPRI, which is encapsulated in Ethernet using IEEE 1914.3 RoE. There are different RoE mappings possible with varying degrees of complexity and bandwidth efficiency. The simplest to implement and possibly the ones with the best complexity-efficiency tradeoff are structure-agnostic RoE mappings, including tunneling mode and line-code aware mode that enable multi-vendor interoperability and the convergence of all CPRI types onto a single Ethernet network. While these have little or no visibility into the CPRI streams, thus limiting the benefit to CPRI bandwidth efficiency, they are the easiest to implement. Other structure-aware mappings have greater visibility into the CPRI structure to improve bandwidth efficiency, though at greater cost and complexity. The legacy CPRI convergence is critical because the embedded base of radios to date is CPRI-based. Operators cannot afford to replace these existing radio access networks as they adopt eCPRI for new 5G rollouts.

Finally, a truly converged mobile network extends beyond fronthaul to include midhaul and backhaul. A physical macrocell site may serve as a fronthaul location for small cell traffic, but also a midhaul/backhaul location for other radios. TSN for Fronthaul-equipped switches supports a unified network architecture that can accommodate traffic from all segments. As C-RANs evolve to cloud RANs (with virtualized network functions), fronthaul, midhaul, and backhaul delineations will no longer be physically fixed. Compute, storage, and applications will place additional flexibility requirements on underlay transport networks.

Figure 4 depicts a converged fronthaul network using IEEE 802.1CM TSN switches.

Figure 4: Converged Mobile Transport with TSN for Fronthaul. Source: Heavy Reading, 2020
Figure 4: Converged Mobile Transport with TSN for Fronthaul. Source: Heavy Reading, 2020

High Precision Synchronization

Legacy CPRI is a synchronous protocol in which synchronization is inherent to the protocol itself. One big change with eCPRI is that synchronization information is not transmitted within the new fronthaul protocol. Thus, while the transition from CPRI to eCPRI reduces the bandwidth demand on the fronthaul link (delivering a roughly 10x reduction in fronthaul bandwidth), it also transfers the synchronization function from the protocol itself to an outside synchronization source.

Synchronization information can be distributed by satellite systems or by the network. Historically, in some countries (including the U.S.), synchronization has been provided by satellite systems that deliver timing from satellites directly to cell sites. With 5G cell site densification, many operators have concluded that satellite alone will not be sufficient – due to costs, availability, and reliability concerns. Network-based synchronization is emerging as a means of distributing synchronization. In other regions, including much of Europe, network-based synchronization is already prevalent using the IEEE 1588v2 Precision Time Protocol.

Significantly, while much of 4G use FDD radio technology (requiring only phase synchronization), 5G makes much greater use of TDD and requires frequency, time, and phase synchronization. Many legacy backhaul networks support frequency synchronization only and must be upgraded to support time and phase. Additionally, performance-boosting techniques, such as carrier aggregation and coordinated multipoint, introduce high precision timing accuracy requirements, as these require close coordination among radios within a cluster.

IEEE 802.1CM and its upcoming IEEE 802.1CMde amendment specify various categories of absolute (concerning a common reference ) and relative (concerning other elements in a cluster) synchronization accuracy for implementation by TSN equipment. The requirements themselves account for differences among specific mobile applications and come from the eCPRI Transport Network Requirements Specification. A simple summary of the categories of importance to transport networks and example timing requirements applicable to Ethernet bridged networks is shown below:

  • Category A: 60 ns-70 ns maximum relative time error
  • Category B: 100 ns-200 ns maximum relative time error
  • Category C: 1.1 μs maximum absolute time error

Categories A and B are relevant for 4G and 5G carrier aggregation radio access technologies (RATs) used between two cooperating RUs. Category C is relevant for TDD RATs.

Important ITU-T timing and synchronization work are referenced in the IEEE 802.1CM standard. Specifically, the ITU-T G.8273.2 standard defines different classes for the time error accuracy/performance targets of telecom boundary clocks/telecom slave clocks implemented in network equipment. These include Classes A, B, and Classes C, D, with A being the least stringent and D the most stringent. (These are different classes than those described above.) For fronthaul, Class C telecom boundary clocks (T-BC) with 30 ns equipment accuracy are needed in the fronthaul segment (between the RU and the DU) to meet RAN and RoE performance requirements. The upcoming IEEE 802.1CMde amendment to IEEE 802.1CM will further refine the timing and synchronization requirements for Ethernet fronthaul.

Ultra-Low Latency

In 5G fronthaul networks, latency requirements are set to account for the following:

  • The needs of applications riding over the 5G networks
  • Network-imposed restrictions from the physical separation of the RU and baseband processing

While ultra-low latency applications such as remote control of critical infrastructure require one-way latency of 1 ms (stringent by 4G standards), HARQ communications requirements typically limit one-way fronthaul transmission latency to 100 μs.

The IEEE 802.1CM standard seeks to reduce latency for the aggregation and switching portions of the fronthaul network. Latency in aggregation is significant because, at an aggregation node, frames may contend for the same port at the same time, leading to buffering (delay) as well as delay variation. Profile A addresses latency in two ways. First, time-sensitive traffic is prioritized over non-time-sensitive traffic, so that the non-time-sensitive traffic gets buffered as time-sensitive traffic receives scheduling precedence. Additionally, Profile A limits maximum frame size to 2,000 octets to minimize the amount of time that time-sensitive traffic must wait when it arrives behind non-time-sensitive traffic.

Profile B goes a step further by adding frame preemption (originally specified in IEEE 802.3br and 802.1Qbu), which allows transmission-in-progress of a non-time-sensitive frame to be interrupted when a time-sensitive frame needs to be transmitted. Once the time-sensitive frame is expressed, the transmission of the non-critical transmission can resume. Profile B with frame preemption has two implications compared to Profile A. First, frame preemption eliminates the need to limit frame size. Second, it assures a bounded latency (marked by tightly controlled delay variation) that is not possible with Profile A. This is important when mixing fronthaul with non-fronthaul services.

While frame preemption provides performance gains through bounded latency, it does require more sophisticated network equipment, which is why the IEEE standard positions Profile A as the simpler option of the two. The real value of frame preemption is its ability to deliver a bounded latency, which cannot be guaranteed with prioritization alone.

Carrier OA&M and High Resiliency/Availability

As a carrier-grade technology, TSN for Fronthaul is designed to operate over a carrier Ethernet network with both carrier-grade operations, administration, and management (OA&M) and carrier-grade resiliency/availability. The considerable body of Ethernet OA&M work done in the early years of carrier Ethernet, when Ethernet was first introduced into operator networks as an alternative to the private line, is key to assuring the performance allowed by TSN for Fronthaul. The IEEE 802.1ag standard specifies fault management protocols in 802.1 bridges and bridged networks. ITU-T Y.1731 specifies additional fault management as well as performance monitoring protocols in Ethernet transport networks. MEF 35.1, 30.1, and 17 are MEF standards governing Ethernet service OA&M. These standards are complementary to IEEE 802.1CM and are essential for carrier-grade Ethernet transport networks.

Future and Industry Next Steps

The standardization of standalone 5G in Release 16 enables the full spectrum of advanced 5G applications and services. For mass-market adoption, standalone 5G also requires the new transport technologies and architectures that have been discussed for several years but have not yet been fully adopted. IEEE 802.1CM TSN for Fronthaul is particularly promising as a packetized fronthaul option for C-RANs. As a packet technology, TSN for Fronthaul is an ideal transport layer for eCPRI, which was developed to work on packet networks.

Equally significant, however, is the ability for IEEE 802.1CM-based networks to provide a converged transport network that extends beyond fronthaul to include midhaul and backhaul segments and to include legacy traffic (such as 4G services and CPRI protocol) as well as new 5G. Interworking between 4G and 5G will be particularly important because 4G services will remain active for many years to come. To achieve a mass-market 5G scale with profitability, a converged 4G and 5G transport infrastructure will be crucial.

Source: NOKIA