Enabling 5G X-Haul with Deterministic Ethernet - A TransPacket whitepaper
1 like238 views
This white paper discusses the concept of Ethernet X-Haul as a transport network for 5G, focusing on the necessity for deterministic Ethernet to meet the stringent requirements of mobile communication. It introduces a patented technology called Fusion that optimizes Ethernet performance by minimizing latency and packet delay variation while enabling the integration of various transport services. The paper also highlights two main use cases, Ethernet Crosshaul and indoor coverage, showcasing how Ethernet can efficiently support both fronthaul and backhaul applications.
Enabling 5G X-Haul with Deterministic Ethernet - A TransPacket whitepaper
- 1. © TransPacket AS WHITE PAPER 5G Ethernet X-Haul June 2018 ______________________________________ Executive Summary The concept of Ethernet X-Haul is being discussed extensively in the mobile industry in connection with 5G. The idea is to have an Ethernet based converged transport network serving multiple purposes including fronthaul and backhaul. This white paper presents the RAN architectures under consideration for 5G, and their consequences in terms of requirements for the transport network. It further describes how an innovative Ethernet scheduling mechanism is required to support deterministic Ethernet, and to fully achieve an 5G Ethernet X-Haul. The white paper is also introducing two use cases, namely Ethernet Crosshaul, and Indoor Coverage, which demonstrate the added value of deterministic Ethernet for mobile transport applications. 1. 2.
- 2. TransPacket White Paper – 5G Ethernet X-Haul 2 © TransPacket AS 1. Introduction 5G is the latest mobile generation that covers all the technologies required to meet ITU IMT-2020 requirements, where three main classes of use cases have been identified: enhanced mobile broadband (EMBB) offering higher throughput, massive machine type communication (MMTC) offering high mobility and high connection density, and ultra-reliable low latency communication (URLLC) offering (near) real- time support. To be able to serve the wide variety of applications coming from those three classes of use cases, a change in the radio access network (RAN) is required, all the way from the air interface to the packet core and the transport network. This white paper focuses on the RAN transport network, and in particular on the mobile fronthaul that is a key enabler for the mobile transport network for 5G. There is a consensus in the industry that fronthaul network will be based on Ethernet, and several forums and standardization bodies, e.g. CPRI and IEEE 1914, are redefining the fronthaul network towards a packet-based architecture. Even though the goal is to define a multipoint-to-multipoint Ethernet switched network for fronthaul, most of the existing studies, prototypes, and pre-commercial deployments include a “network” that is implemented as a dedicated point-to-point Ethernet link. Some of the main technical reasons for this first stage solution are the well- known quality of service (QoS) issues connected to Ethernet switching, namely latency, latency variation and time and frequency synchronization. This white paper highlights how available technology from TransPacket adding deterministic QoS mechanisms to Ethernet can be used in order to design a truly switched Ethernet fronthaul network meeting the strict requirements of latency, latency variation, and synchronization of fronthaul traffic. The ability to deploy fronthaul over a switched Ethernet infrastructure offers the possibility to carry fronthaul services together with midhaul, backhaul, and fixed access services over the same Ethernet transport infrastructure, achieving what is commonly referred to as an X-Haul network. 2. RAN Architecture and Transport Network Requirements Figure 1 illustrates main RAN architectures. It shows the introduction of the fronthaul network, originally described as a centralized-RAN (C-RAN), where Remote Radio Heads (RRH) are connected to a Baseband Unit (BBU) pool in a Central Office or Data Centre through high bandwidth transport links running typically the Common Public Radio Interface (CPRI). Because of the challenging capacity and latency requirements, CPRI-based fronthaul is most often deployed over dark fiber or WDM solutions. Those solutions for fronthaul meet the performance requirements of today’s 4G/LTE networks, but for many Mobile Network Operators (MNOs), these are too expensive and inefficient for 5G. Figure 1 - RAN Architectures The third architecture, referred to as Cloud-RAN (C-RAN), is based on a) enabling different functional splits by centralizing higher layers of baseband processing functionalities while maintaining lower layers of baseband functions in the RRHs, b) virtualizing higher layers of baseband functions to be run on general
- 3. TransPacket White Paper – 5G Ethernet X-Haul 3 © TransPacket AS purpose servers in remote data centers, and c) an Ethernet-based fronthaul. Both the functional split and the Ethernet fronthaul impact the RAN transport network, and are addressed further in the next sections. 2.1 Functional Split In order to ease the strict requirements in terms of bandwidth and latency on the fronthaul, a number of alternative functional splits of the baseband functionalities have been proposed and discussed by the industry. Eight options have been studied – where option 8 aligns with what is known as CPRI. There is no one-fits-all split, and the “optimal” split depends on a number of technical and business parameters like network topology, availability of fibers, number of users, volumes of services, etc. In order to be able to address this variety of scenarios, and at the same time to limit the number of options to reduce complexity and deployment cost, 3GPP has proposed to support two splits: a high-level split, and a low- level split. The high-level split means that real time processing is performed by the Remote Radio Head, relaxing the requirements of bandwidth and latency, but making inter RRHs coordination in connection with e.g. Coordinated Multi-Point (CoMP) more challenging. The low-level split implies keeping low-complexity and low-cost RHHs, and centralizing more functionality in the BBU. But unlike the traditional (CPRI) functional split, the data throughput is variable and proportional to the user data rate. The interface for the low-level split is time sensitive, and strict latency control is required. IEEE 1914 has defined a model for the functional splits by introducing three logical block functions: Radio Unit (RU), Distributed Unit (DU), and Centralized Unit (CU). Those functions can be distributed between the antenna site and the central office site as illustrated in Figure 2. The placement of the building block defines different networks, namely fronthaul, midhaul and backhaul, with different requirements in terms of latency, capacity, and maximum reach. Figure 2 - RAN Functional Splits
- 4. TransPacket White Paper – 5G Ethernet X-Haul 4 © TransPacket AS The fronthaul network has a very strict latency requirement (under 100 µs) driven by the infrastructure1 . But it is important to remember that even if a high-level split is used (implying a more relaxed latency requirement on the transport network from the infrastructure perspective), 5G is promising to serve applications with very low latency (down to 1 ms), and these time-sensitive applications will still demand a low latency transport network. In addition, as mentioned earlier, there will not be a single split, and the transport network will have to support a mix of splits with both time-sensitive and less time-sensitive/time-insensitive traffic. This calls for Deterministic Ethernet in the mobile transport network with the capability to support a mix of traffic carrying fronthaul, midhaul, and backhaul services. 5G is foreseen to address a wide range of applications, with, to some extent, conflicting requirements. As MNOs cannot afford to build several dedicated transport networks to carry those services, there is a need for supporting all services over a common infrastructure. This calls for Network Slicing, which enables to dedicate a “slice” of the physical network to deliver the required KPIs for a set of applications while being completely isolated from (i.e. not affected by) other slices. 2.2 Ethernet Fronthaul Ethernet fronthaul requires encapsulation of the radio signal in an Ethernet frame. There are two competing standards for encapsulating fronthaul traffic in an Ethernet frame: a) eCPRI from the CPRI consortium, and b) Radio over Ethernet (RoE) from the IEEE 1914.3 Working Group. While the encapsulations themselves are different, the principle is similar in both cases, and they both require low delay and low packet delay variation (PDV). Indeed, to ensure correct processing of the encapsulated data, a steady stream without PDV is required. In practice, the PDV can be removed by a playout buffer at the receiver side with the minimum buffer size equal to or greater than the peak PDV. However, this requires knowledge of the maximum delay and peak PDV to dimension the network. The challenge with Ethernet is that packet delay and delay variation is dependent on traffic load. The TSN standard for fronthaul (IEEE 802.1CM) includes a mechanism (IEEE Qbu) to preempt low priority traffic in favor of high priority fronthaul traffic but does not include deterministic scheduling of multiple fronthaul streams. The current TSN scheduling mechanisms (IEEE 802.1Qbv, IEEE 802.1Qch) targeting industrial automation applications were not deemed suitable for fronthaul, and have not been included in the IEEE 802.1CM standard. There is, therefore, a need for additional scheduling mechanisms in order to support a deterministic Ethernet-based network for fronthaul. Deterministic scheduling is required for aggregation purposes. In 5G, the need for more capacity and the use of high-frequency bands (and their smaller coverage) will drive densification. As building new macro sites is expensive and complex in terms of both space requirements and environmental impact, densification will lead to a significant increase of small cells. It is projected that in busy areas, e.g. downtown of major cities, there will be small cells at every block to support 5G services. This clearly creates a need for aggregation and Ethernet switching as it is unlikely that connecting every single small cell via dedicated fiber to the vBBU is a scalable and economic model. 1 For the fronthaul, the strict latency (under 100 µs) is driven by the HARQ protocol.
- 5. TransPacket White Paper – 5G Ethernet X-Haul 5 © TransPacket AS 3. FUSION TransPacket has developed and patented a technology, FUSION, which enables MNOs to design and deploy Ethernet X-haul networks. The focus of the technology is to minimize the packet delay and delay variation even in the presence of multiple traffic sources. 3.1 Concept The key concept relies on preserving packet gaps for priority traffic, and to take advantage of the inter- packet gaps between high priority packets to transport low priority packets. FUSION also enables the aggregation and convergence of multiple fronthaul, midhaul, and backhaul streams in the same network. FUSION achieves ultra-low PDV in the range on nanoseconds for fronthaul, and enables the multiplexing of midhaul and backhaul traffic over the same transport network. These characteristics are achieved by implementing the following scheduling features: • Deterministic aggregation: the capability to aggregate several fronthaul streams with low and fixed delay. • Deterministic priority: the capability to statistically multiplex lower priority traffic, e.g. midhaul and backhaul traffic, with no impact on fronthaul streams 3.2 How it works The FUSION scheduler adds a fixed delay to the fronthaul traffic at each node. This enables the look ahead for identifying gaps within this fronthaul stream, and filling/inserting less delay sensitive backhaul/midhaul packets only in fitting gaps. Theoretically, no PDV is induced on the fronthaul stream of packets. Considering backhaul traffic on 100Gb/s links, with MTU of 1500 bytes, the delay is equal to 144 ns, and corresponds to 122 ns, the transmission time of a lower priority packet MTU at the transport interface plus 20 ns, the fixed processing time. While all packets experience this delay, i.e. even if there is no backhaul packet to be transmitted, this enables a fixed PDV per node which simplifies the network and PDV buffer dimensioning challenges for the carrier. FUSION scheduling provides a clear advantage regarding fronthaul traffic performance isolation: delay and delay variation of the fronthaul traffic streams are independent of the number of midhaul/backhaul traffic streams, and of the number of transport switching nodes. The deterministic priority enables to add backhaul traffic at any point in the future. The small delay added by FUSION comes with the advantage of a seamless network upgrade/change, ensuring that the fronthaul peak PDV does not change. The PDV playout buffer size can thus be dimensioned for the maximum number of fronthaul streams that the network shall aggregate. In addition, FUSION can also support a PTP stream with ultra-low PDV enabling accurate synchronization of the mobile nodes in a fronthaul network. 3.3 Key features • Low and Ultra-low Latency FUSION transfers high priority streams unaltered, including gaps, and is, therefore, able to achieve low latency with ultra-low Packet Delay Variation (PDV) for those streams. • Multi-service support FUSION transfers lower priority by using the gaps within the high priority streams. FUSION is therefore able to support both high- and low-priority streams simultaneously, while making sure low priority streams do no impact performance (loss, delay, and PDV) of the priority streams, achieving therefore “soft” slicing of the Ethernet transport network • Network Slicing and Hard isolation FUSION is able to aggregate several high priority streams while conserving their performance (in terms of loss, delay and delay variation) by creating a “virtual wavelength” per high priority stream. It means that the high priority streams are isolated from each other’s, and FUSION achieves “hard” slicing and QoS isolation of the Ethernet transport network.
- 6. TransPacket White Paper – 5G Ethernet X-Haul 6 © TransPacket AS 4. Deployment Scenarios 4.1 Ethernet Crosshaul One of the major concerns of mobile operators is to keep the Ethernet fronthaul simple, and to ensure that the support of deterministic behavior does not lead to a significant increase of their operational cost due the complexity of configuration of mechanisms that depend on too many factors, e.g. topology, load, traffic type, and traffic variation. In IEEE 1914.1, there is an on-going discussion (initiated by MNOs) about defining classes of Ethernet transport nodes for fronthaul, associated with a given upper value for processing time. The objective is to have a fixed value, independent of traffic type and load, in order to simplify the network design, and network upgrade/change. Providing fixed delay Ethernet switching for time-sensitive traffic (fronthaul) is one of FUSION key added value. In addition, the ability to add/drop fronthaul and backhaul/midhaul traffic to existing deployment without impacting the deployed fronthaul services is another key added value of the FUSION scheduler, which therefore addresses the MNO requirements for simplicity, and ease of design. Figure 3 presents the use case for Ethernet transport of fronthaul and backhaul, also known as crosshaul, where a converged transport network is used to integrate both fronthaul and backhaul together with internet access services for both residential and business subscribers. It illustrates how FUSION switching and scheduling IP Core technology is able to support deterministic Ethernet as required by a fully converged mobile transport network. Using FUSION IP Cores, fronthaul traffic from e.g. RoE and eCPRI streams can be aggregated with low and bounded delay, and can be switched with fixed and low delay through the network. The spare capacity in the path is then utilized for lower priority traffic such as midhaul and backhaul, without affecting the timing of the fronthaul traffic streams. Figure 3 - Ethernet Crosshaul 4.2 Indoor Coverage In the recent years, indoor coverage has become a challenge for many MNOs. As 80% of all mobile communications are made indoor, most mobile customers, and especially business mobile customers, expect full indoor coverage. New building rules and new materials have made wall penetration more difficult, and the use of new spectrum (higher frequency) in 5G will make indoor coverage even more challenging. There is, therefore, a need to deploy indoor specific solutions to make sure that indoor coverage is indeed acceptable. The existing solutions for indoor coverage have been either Distributed Antenna Systems (DAS) used for large premises such as stadium or shopping mall, or Small Cells for smaller locations. DAS often requires a dedicated infrastructure, and the costs and complexity of the installation do not scale down enough for
- 7. TransPacket White Paper – 5G Ethernet X-Haul 7 © TransPacket AS smaller premises. Ethernet-based Small Cells offer the possibility to re-use existing LAN (Ethernet) infrastructure and to minimize CAPEX investment. FUSION takes this model one step further by allowing MNOs to deploy simpler and more cost-effective radio units (implementing a low split), and having either a combined DU/CU in the basement, or a DU in the basement, and a CU at a remote location serving several locations. It provides the opportunity to share a single Ethernet infrastructure within a building to support both fixed and mobile networks reducing OpEx even further. As illustrated in Figure 4, hard slicing support enabled by FUSION allows several MNOs to share the infrastructure. This opens up for new business models, and therefore a broader market, where the property owner can share the investment, the risk, and the benefits with one or several MNOs. Supporting multi-operators on the same fronthaul infrastructure is an enabler for new businesses in indoor coverage markets. Figure 4 - Indoor Coverage 5. Summary Ethernet is already the de facto standard for backhaul and it is becoming the solution of choice for fronthaul as well. While Ethernet is used as a switched technology in the backhaul, Ethernet usage in the fronthaul has been limited to dedicated point-to-point connections. This is due to Ethernet well-known QoS shortcomings in terms of latency, latency variation, and synchronization. FUSION IP Cores from TransPacket constitute the missing building block, unique in the market, which enables equipment provider to design and build Ethernet equipment to support an Ethernet switched X-Haul targeting the 5G transport network and the new 5G use cases. Yet, as usual in the telecom industry, there will be a mix of mobile technologies deployed at the same time requiring transport capabilities for backhaul, midhaul, legacy fronthaul (CPRI, OBAI), and Ethernet fronthaul. That is the reality for MNOs that need to make sure it all interoperates as seamlessly as possible, and with the highest performance possible. FUSION is a key technology that enables a truly converged deterministic Ethernet infrastructure for backhaul, midhaul, fronthaul and fixed services transport.