5G Mobile Core Networks: Migration From NSA to SA 5G Core

Transcription

1 Independent market research and competitive analysis of next-generation business and technology solutions for service providers and vendors 5G Mobile Core Networks: Migration From NSA to SA 5G Core A Heavy Reading white paper produced for Huawei AUTHOR: GABRIEL BROWN, PRINCIPAL ANALYST, HEAVY READING

2 4G/5G CORE NETWORK STRATEGIES Operators worldwide are preparing to deploy 5G networks to improve the economics of mobile broadband and to introduce new capabilities and service types. The core network, which controls authentication, session management, mobility, quality of service (QoS) and more, is critical to operating 5G and to enabling advanced services, such as network slicing and converged fixed/mobile access. It is this role the core plays in services and monetization that make it critical to a mobile operator's 5G business strategy. For early 5G deployments, many operators are expected to deploy 5G radio access networks (RANs) in non-standalone (NSA) mode on a host Long Term Evolution (LTE) network. Using an LTE radio for over-the-air signaling, and a 4G core network to anchor and manage sessions, offers a fast, practical way to introduce 5G into commercial service. Over time, to meet new service requirements, operators will introduce dedicated 5G core networks to support 5G in standalone (SA) mode. This 5G core will ultimately serve as a common core for evolved LTE (elte) and 5G RANs, as well as potentially for wireline and WiFi access. This white paper discusses this transition from 5G NSA to 5G SA. It addresses how operators can deploy 5G RAN on an upgraded Evolved Packet Core (EPC) and evolve over time to a next-generation, cloud-native 5G core. It investigates the numerous potential migration options and identifies the most attractive scenarios, according to factors such as the operator's service vision, coverage strategy and expected uptake of 5G devices. This paper makes the case that a common, integrated 4G/5G core will be attractive for a majority of operators in the medium term. It also argues that modern cloud technologies and the new 5G system architecture make it economical to operate multiple virtual core network instances on common infrastructure. Non-Standalone & Standalone 5G The reference document for the 5G core is Technical Specification , "System Architecture for the 5G System." This document defines the core architecture, the functional elements and the high-level interfaces between them. The specification of the 5G core is informed by modern networking concepts: It will be cloud-native, distributed, state-efficient and will use service-based interfaces. 3GPP Release 15 the first 5G standards release includes "bare bones" 5G core specifications. The full 5G core specifications, compatible products and deployments will come in later releases. In the meantime, to accelerate launch, 5G RAN can be deployed in NSA mode in a host LTE network using a 4G-EPC. This is attractive to operators that already have large, capable LTE networks. Over time, operators will migrate to 5G in SA mode using a 5G core. Broadly speaking, there are three major factors that drive NSA to SA migration strategies: 5G radio coverage: In the early phases, LTE coverage will exceed 5G coverage by a significant margin. For example, where 5G mmwave radio is deployed, the difference relative to low-band LTE will be extreme. In these scenarios, the use of the LTE control plane is attractive. Over time, 5G coverage will increase, changing the calculation especially when low-band 5G is deployed, either in fresh spectrum (e.g., 600 MHz or 700 MHz) or in re-farmed bands. Contiguous city-level 5G coverage could also be achieved reasonably quickly using mid-band spectrum (e.g., 3.5 GHz or 4.5 GHz). HEAVY READING HUAWEI 5G MOBILE CORE NETWORK MIGRATION STRATGIES 2

3 Device capabilities: Both NSA and SA have important dependencies on devices, and the rate at which these capabilities are supported will partly determine the rate at which operators can migrate to a 5G core. These dependencies are both frequencyand protocol-specific. Even if the operator deploys enhanced 5G coverage and has devices that support SA, it will still need to support the installed base of NSA devices from the early 5G phase (likely to be premium-tier devices with several years of life). The operator will also need to support a large base of 4G-only devices for a long time. Services strategy: To deploy advanced 5G services, such as end-to-end network slicing, a 5G core and 5G over-the-air signaling will be needed. Certain use cases, particularly where premises-based or venue-based networks are concerned, will likely need this capability sooner than the wide-area network (WAN) e.g., a 5G wireless network for industrial Internet of Things (IoT) will need a local 5G core to meet demanding latency and jitter requirements. As 5G coverage improves, there will be a stronger case for pursuing advanced services in the WAN, which will accelerate SA mode. The migration from NSA to SA is shown in Figure 1. Although the chart looks simple, it is a complicated transition, with lots of factors at play and several migration options between these two states. Figure 1: Migration for NSA to SA Mode 5G Networks ANALYSIS OF 5G CORE OPTIONS 3GPP TR , "Study on new radio access technology: Radio access architecture and interfaces," identified a range of options for the migration from NSA to SA. The fact that this was a RAN initiative within 3GPP shows how important radio is to the core network strategy. It also underlines that the core should be thought of as part of a broader 5G system architecture. Outline & Compare Options 2-7 There are least 10 options for the migration from 5G RAN deployed using NSA to a future state where a 5G core is in operation. These options were developed to inform the specification process. They are shown in Figure 2, with the leading candidates outlined in red boxes. HEAVY READING HUAWEI 5G MOBILE CORE NETWORK MIGRATION STRATGIES 3

4 Figure 2: Migration Options for NSA to SA 5G Networks The reason to develop so many options 10 is obviously too many was to help inform the specification process. An analysis that considers RAN, device and service strategies leads us to identify Option 3x, Option 2 and Options 4/7x as most attractive, and we expect these to be developed further in Release 16. The other options can be discounted for various reasons, as follows: Option 4a is redundant in most cases. If the user has a 5G device that supports 5G control plane and user plane over the air, then Option 2 using a standalone core may HEAVY READING HUAWEI 5G MOBILE CORE NETWORK MIGRATION STRATGIES 4

5 be appropriate. If that operator has also upgraded to elte, then Option 4 would be preferable to having dual core network connections. Option 3 and Option 7 both have a single connection to the core network via the enb. This is likely to meet a user-plane bottleneck, as 5G traffic will be routed through the 4G base station, which may be older. For this reason, they can both be discounted. Mobility in Option 3a and Option 7a will be complicated due to dual core network connections. This will scale from a user-plane perspective, but handover between LTE and 5G will be more challenging because there is no Xn interface between RANs. Option 5 is effectively redundant. An operator that has both an elte RAN and a 5G core would typically also have a 5G RAN. Options 2/3x/4/7x Are Most Attractive This leaves Options 2, 3x and 4/7x as most attractive, for the following reasons: Option 3X is the optimal NSA configuration for operators that want to provide integrated services with robust LTE-5G mobility across different frequencies and coverage zones. In this case, over-the-air control plane is supported by the LTE radio. The gnb has S1-U interface to EPC and an X2 interface to the enb. 5G NR carries the user-plane traffic, which limits the upgrades needed to the existing LTE RAN and transport network. When NR coverage is poor, the device falls back to LTE user plane. This is an example of EUTRA-NR Dual Connectivity (EN-DC). Option 4 and Option 7x may be adopted where operators have aggressive 5G RAN deployment plans and want to offer integrated LTE/5G services with robust inter-rat mobility. Option 4 is primarily under consideration in China. In both cases, the LTE RAN should be upgraded to elte. These are examples of NG-RAN EUTRA-NR Dual Connectivity (NGEN-DC). At this stage, the 5G core becomes a common packet core. Option 2 allows 5G deployment independent from LTE. It is also an idealized end state (to the extent that such a thing can ever exist). This option is also suited for greenfield 5G, where there is not a requirement for inter-rat handover to LTE, or for operators with excellent 5G coverage that are happy to interwork with LTE at the core level, rather than in the RAN. DEPLOYING "5G-EPC" USING NSA OPTION 3X NSA mode is attractive precisely because it is less disruptive than deploying a new 5G core. Operators generally have well-functioning EPCs and want to re-use those assets and operational processes for 5G RAN. "5G-EPC" does, however, require investment to operate at 5G scale and operators must make some important design choices. Should 5G-EPC be deployed as an overlay discrete from 4G-EPC? Or should it be combined into a common 5G/4G pool? How should metering and charging work? What should be the voice strategy? Hybrid Pool for 4G/5G Access Perhaps the simplest solution is to pool core network functions for 4G and 5G accesses. This means the same subscriber databases and profiles are used for authentication and policy, the same MME pool for mobility and the same gateway pools for user-plane processing and forwarding. There is obvious synergy to this to model because: HEAVY READING HUAWEI 5G MOBILE CORE NETWORK MIGRATION STRATGIES 5

6 In the first instance, hybrid pool minimizes the number of elements that must be deployed and managed, and so simplifies network operation. This includes peripheral functions such as related to monitoring and service assurance. This is important to the overall cost-of-production and cost-per-bit economics of 5G in NSA mode. A common gateway and policy infrastructure will also simplify service management. The same services will be available to all subscribers, thereby simplifying marketing of 5G, making it transparent to the customer. If desired, differentiated 5G services could be offered on this infrastructure using policy control. Using a common P-GW enables re-use of metering and charging infrastructure and does not require new integration to the BSS. 5G service plans can be supported via policy, but the actual metering and charging integration remains the same. This is clearly an attractive option and will likely be pursued by many operators. However, 5G NR in wideband spectrum is expected to generate significantly greater traffic volumes than LTE. To meet this demand operators will need to invest in gateway capacity and use control- and user-plane separation (CUPS) for scalability. They will also need investment in additional backhaul and metro transport. DECOR for 5G-EPC The 3GPP standards make provisions for the creation of dedicated EPC networks connected to a common RAN. This specification is known as Dedicated Core Network (DECOR) and was intended for applications such as virtual private LTE networks and enterprise IoT. It is conceptually similar to the Multi-Operator Core network (MoCN) specifications, where different PLMNs are used for network sharing and MVNOs. In principle, virtualization and cloud technologies make it much easier and more cost-effective to deploy dedicated gateways. The main advantage of this approach is that the dedicated core network instance can be dimensioned according to the requirements of the use case. In LTE, this typically means services such as connected car or IoT, which have transaction density, mobility and throughput profiles that are very different from standard smartphone services. An example 5G service where DECOR could make sense is fixed-wireless broadband, so that appropriate policies, traffic management and charging can be applied to the residential service that may not apply to a smartphone service. This model keeps fixed broadband traffic segmented from regular mobile broadband traffic and allows for independent scaling of the different service types. CUPS & Distributed EPC Control- and user-plane separation (CUPS) allows for independent scaling of the S/P-GW and will be important to meet the extra traffic generated by 5G RAN. Strictly speaking, CUPS is a 4G enhancement conceived to support advanced LTE RANs; however, it has emerged as a defining feature of a "5G-EPC" and is critical to NSA mode. Operators can benefit from CUPS deployed in a central location i.e., where their existing EPC is located due to independent scaling of the 5G user plane. However, a feature of CUPS is to enable a distributed user plane. Figure 3 shows how this works conceptually with gateway deployment closer to the edge, probably occurring in phases. HEAVY READING HUAWEI 5G MOBILE CORE NETWORK MIGRATION STRATGIES 6

7 Figure 3: CUPS Evolution to 5G Architecture Several operators are deploying distributed "5G-EPC" to support 5G launch. City-level and regional deployments look appropriate in the first phase, with greater distribution in later phases. As a rule of thumb, a tenfold increase in gateways, over time, relative to 4G EPC, looks appropriate for mobile broadband services in the WAN. The CUPS model has also been adopted formally in the 5G Core specifications, which means a virtual U-PGW has a straightforward upgrade path to UPF in the 5G core. NSA Voice Strategy The 3GPP voice strategy for 5G in NSA mode is to reuse voice over LTE (VoLTE). This is logical in the sense that most operators have recently deployed VoLTE and IP Multimedia Subsystem (IMS) and do not have the appetite for another upgrade. Where the coverage of LTE is superior to 5G NR, it also makes sense to use LTE to make or receive calls, even when the user is connected to NR. This is sometimes called "EPS fallback" and is analogous to CS fallback from LTE to 2G/3G. Over time, the new 5G core will connect to a common IMS (or IMS-like) core for real-time communications. At this stage, the new 5G QoS and bearer model will ensure the quality of these sessions. It is an open question how long IMS will be used; it could be that a new real-time communications subsystem is introduced for 5G. DEPLOYING A COMMON 4G/5G CORE Operators generally prefer to operate common core infrastructure for multiple access technologies. This reduces cost of operations by limiting the duplication of design, integration, deployment, maintenance, updates, training, etc. On the service side, a common core makes it simpler for operators to offer a consistent service portfolio. A common core doesn't, however, preclude 5G-only services. It is also possible to support multiple options on a common infrastructure, making the migration to 5G core conceptually simpler. Common 4G/5G Core A hybrid 4G/5G core is a way for operators to migrate investment in EPC to the new 5G core network as 5G subscribers and traffic grow. Figure 4 shows this logical core network architecture. The red section highlights common user-plane pool; the grey section highlights common control-plane elements. HEAVY READING HUAWEI 5G MOBILE CORE NETWORK MIGRATION STRATGIES 7

8 Figure 4: Integration Points for a Common 4G/5G Core Source: 3GPP, Huawei, Heavy Reading There are three main points of convergence in the common 4G/5G core. These are: Convergent subscriber data: Common subscriber databases for the home subscriber server (HSS) and unified data management (UDM) functions offer a clear benefit in terms of scale, operations and security. Subscriber data systems will continue to evolve, but this is not directly dependent on RAN technologies. For example, the UDM may also serve fixed-access subscribers over time. Convergent control plane: This includes a common PCF/PCRF for policy management. The overall policy framework will evolve to incorporate network slicing in 5G. Also in the control plane are mobility and session management. In this case, the 4G MME and the 5G AMF may not be as easily integrated, and may remain separate. Convergent user plane: A PGW-U will be similar in many respects to a 5G UPF. Both terminate GTP, both apply policy and traffic management, and both route IP traffic. This model also supports common traffic processing environment on the SGi and N6 interfaces, such that common firewalls, Web proxies, etc., can be used. This common core infrastructure approach can support Option 3x, Option 7x/4 and Option 2, and gives operators flexibility to adapt as standards evolve, device compatibility becomes clearer and 5G coverage improves. Migration Path From 3x (NSA) to Option 2 (SA) With a common infrastructure and the options narrowed down, there are two clear migration paths: Either start with Option 3x, upgrade to Option 7x/4 and then migrate to Option 2 (SA); or go directly from Option 3x to Option 2. Both paths are shown in Figure 5. HEAVY READING HUAWEI 5G MOBILE CORE NETWORK MIGRATION STRATGIES 8

9 Figure 5: Migration From EPC to 5G Core The Option 3x > Option 7x/4 > Option 2 migration offers a good solution where the operator wants tight 4G-5G interworking at the RAN level. However, this path increases costs in terms of gnb software upgrades and reconfiguration of transport. Moreover, neither Option 7x nor Option 4 have a committed device chipset roadmap at this stage. In some cases, especially where extensive 5G coverage is deployed, the operator may migrate directly to a 5G core using Option 2. The 5G core and 4G EPC would interwork, while the RANs remain logically sperate, but physically collocated. This is shown and discussed below (see Figure 6). Direct to 5G Core (Option 2) With 4G-EPC Interworking For operators with aggressive 5G coverage plans, there may not be a requirement for tight integration with LTE RAN. Greenfield 5G operators would be the obvious example; however, they are likely to be few in number. However, for operators with excellent LTE networks, there is also a case for standalone 5G core, assuming 5G coverage and 5G SA-mode devices. The advantage of this model is that it reduces complexity, relative to 7x/4, and in principle allows the 5G network to evolve without the legacy, perhaps at a faster pace. Figure 6 shows Option 3x (NSA) used for devices with LTE dependencies and Option 2 (SA) used for 5G-only devices and services that require a 5G core. This is an interesting model that HEAVY READING HUAWEI 5G MOBILE CORE NETWORK MIGRATION STRATGIES 9

10 appears to be gaining momentum due to relative simplicity and expectations for rapid 5G RAN deployment. Each option can be deployed on common infrastructure with shared functions. Figure 6: Standalone Core With Interworking SA Voice Strategy Voice over 5G New Radio (VoNR) is supported in 5G core and RAN. The communications services themselves are supported by the existing IMS core, which essentially means VoLTE and VoNR are different access modes for IMS voice/video communication services. Figure 7 shows evolution from VoLTE to VoNR. Path 1 refers to NSA mode, as discussed above. EPS fallback (EPS FB) is used. Path 2 refers to a 5G direct access to IMS without fallback to LTE. The 5G QoS model will support this, but further specification work is needed. Path 3 refers to IMS service via the 5G core using elte RAT Fallback. In this case, 5G NR does not provide voice/video communication services. When the gnb receives notification of a call, a handover is triggered to redirect the device, via inter-rat handover, to the elte network. A VoeLTE service is then used for the call itself. Figure 7: Voice Evolution From VoLTE to VoNR 5G Source: Huawei, 2018 HEAVY READING HUAWEI 5G MOBILE CORE NETWORK MIGRATION STRATGIES 10

11 Standalone 5G Core for Industrial IoT In some circumstances, a standalone 5G core makes sense in the earliest phases of 5G. One example is for industrial URLLC use cases (factories, warehouses, etc.), where the 5G core is deployed on-premises very close to the 5G RAN to ensure low latency and high reliability. Figure 8 shows how this might apply to robotic motion control in a factory automation use case. This type of network operates largely independently from the wide-area mobile network. Figure 8: Standalone 5G Core for Local-Area URLLC In this example, the network must be locally deployed to meet the 2 ms cycle time required by the application. And even where cycle times are somewhat relaxed for example, 10 ms for control of automated guided vehicle in a factory or warehouse there is still a requirement for on-premises deployment because of the need for reliability, maintainability and control. Next Phases in 5G Core The first phase of 5G core specifications were included on Release 15 and specifically referencing Option 2. This is the start of what will be a long-term evolution. Release 16 specifications will include more complete 5G core, as outlined in Figure 9. (Release 16 also features several study items investigations into new capabilities for potential inclusion in later Releases. The 3GPP has a full list of the SA2 study items with direct impact on the 5G core.) Figure 9: Major Differences in NG Core Features in Phase 1 & Phase 2 Release 15 / Phase 1 Release 16 / Phase 2 No support for connectionless mode, e.g., for IoT; decision made by RAN Connectionless mode to support mmtc services in RAN and core 3GPP Access and untrusted non-3gpp (e.g., WiFi) Converged access types (e.g., fixed access) EPC-like access selection Limited URLLC support Cellular topology; no relay or mesh support New network discovery and selection mechanism More complete URLLC specifications Support for relays and mesh, pending a base station architecture from RAN HEAVY READING HUAWEI 5G MOBILE CORE NETWORK MIGRATION STRATGIES 11