1 The Status of Open Source for 5G 2 The Status of Open Source for 5G TABLE OF CONTENTS Executive Summary . 4 1. Introduction . 4 2. Open Source (OS) Model . 5 2.1 What is Open Source? . 5 2.2 Governance Models Make a Difference. 6 2.3 Standards-based Open Source vs. Open Source implementations . 6 3. 5G Architecture . 7 3.1 5G Core Network . 7 3.1.1 Control Plane (CP) . 8 3.1.2 User Plane (UP) . 10 3.2 5G Radio Network . 10 3.3 Convergence Support for non-3GPP Access . 12 4. OS Applicability in 5G . 13 4.1 5G Infrastructure . 14 4.2 5G Radio Network . 15 4.3 5G Core Network . 15 4.3.1 Control Plane (CP) . 16 4.3.2 User Plane (UP) . 16 4.4 Convergence . 17 4.5 Management various options of dual-connected mode, where both LTE and NR are deployed, are not provided; also note that an O-RAN 7 based architecture is being developed by a new alliance with participation from major service providers. A key aspect of 5G RAN is to disaggregate eNB/gNB into multiple modules (DU, CU-CP, CU-UP, etc.), see figure 3.2 and define standard interfaces between them (e.g. New Open Front haul, A1-o (OAM), A1-c (control) E1, E2, F1 etc.) to support interoperability between various components coming from different vendors. This is being done by O-RAN in parallel to 3GPP. Additionally, separating control plane (CU-CP) and user plane (CU-UP) of new radio allows operator to implement advanced area optimization applications (e.g. per UE level optimization, Mobility management, radio connection management, load balancing, etc.). Ultimate goal is to allow operators to leverage open source community to implement best in class area optimizations applications in a vendor neutral CU-CP that is flexible enough to support plug-n-play optimization algorithms. Some of these concepts are being introduced to the larger community in 2019 and would require performance objectives to be tested and proven in follow-on work. 7 o-ran/resources/ 11 The Status of Open Source for 5G Figure 3.2. O-RAN Reference Architecture 8 The primary functional element in a basic 5G radio network is the disaggregated element of gNB, CU-CP, which is a radio node responsible for connecting the UE to the 5G core. It terminates user and control plane traffic from a UE sent over the 5G air link. It then interfaces the control and user plane of the radio network to the 5G core. It also transparently passes signaling communication Non-Access Stratum (NAS) between the UE and the 5G core. A gNB can be further split into two sub modules. gNB Central Unit (gNB-CU) A gNB-CU is part of a gNB. It is a logical element that is home to higher layer radio protocols that can be placed in a more centralized manner. A gNB-CU controls the operation of one or more gNB-DUs and also communicates with other gNBs for tasks such as handovers. It has further CP and UP components as shown in Figure 3.2. gNB Distributed Unit (gNB-DU): A gNB-DU is part of a gNB. It is a logical element of home lower layer radio protocols that can be placed in a more distributed manner. One or more gNB-DUs are controlled by a gNB-CU. As shown in Figure 3.3, a gNB-CU, along with one or more gNB-DUs, form a logical gNB. The interface connecting gNB-DU and gNB-CU is a point-to-point connection that carries both signaling and data traffic. 8 o-ran/resources 12 The Status of Open Source for 5G Figure 3.3. Simplified 5G Radio Network with NR. 3.3 CONVERGENCE SUPPORT FOR NON-3GPP ACCESS 3GPP specifications have historically supported non-3GPP access types as part of 4G EPC. Examples of non-3GPP access types include Wi-Fi and wireline. Traditionally, such support in 3GPP has been categorized into two types: untrusted and trusted. Untrusted can simply be described as an over-the-top model where it is assumed that local connectivity via non-3GPP access is available somehow and a secure tunnel is established between the UE and 3GPP network using 3GPP credentials. For the untrusted model, 3GPP can develop solutions without external dependencies. In Release 15, as part of 5G work, only the untrusted model is supported. In Release 16, 3GPP is collaborating with the Broadband Forum to support the trusted model in 5G. Figure 3.4 shows a high-level architecture for the untrusted model that is supported in Release 15. Only key elements of the 5G core are shown. For example, a multi-access UE can be connected simultaneously to 5G NR and Wi-Fi/wireline. In this case, multiple data and signaling paths will exist from the UE. 13 The Status of Open Source for 5G Figure 3.4. Support for Convergence in Release 15. N3IWF The key element of the untrusted access model is a new function named “Non-3GPP Inter-Working Function” (N3IWF). The UE uses N3IWF to connect to the 5G core over a non-3GPP access layer. N3IWF terminates the security tunnel from the UE side and terminates signaling and data plane from 5G core functional entities. The UE is assumed to support the 5G signaling plane (NAS). N3IWF carries both NAS and user plane data between the UE and 5G core functions. 4. OS APPLICABILITY IN 5G With the emergence of network softwarization, open source software is going to be critical, as described in the following subsections, to 5G and will play an important role in the development of 5G networks. Open source projects have a global developer community for the task of solving technical challenges and expanding access to technical capabilities. The community model of open source development identifies and responds to user needs more rapidly than perhaps the traditional standards world. Open source could help operators find interoperable solutions, encourage innovation, improve quality and security and contribute to the community. The open source approach also helps vendors free up resources to pursue value-added products/services, improve quality and security and contribute to the community. With the open source movement quickly gaining momentum, there are numerous open source projects across multiple domains of infrastructure, management, control, access and core. Determining the applicability of open source to an appropriate domain can be overwhelming. Section 4 covers the aspects to be considered when determining the applicability of open source across different layers of the 5G network. 14 The Status of Open Source for 5G 4.1 5G INFRASTRUCTURE In order to provide massive amounts of bandwidth to a massive number of devices, there is a need to transform the network to be able to scale up and be agile while reducing cost. Network disaggregation with separation of user and control plane, separating out the network operating system from the underlying hardware, and use of general-purpose processing platforms is the key to creating networks that are massively scalable, agile and inexpensive. Disaggregated hardware provides high performance at lower costs via approaches such as specialization of tasks (for example, servers designed for packet processing) or conformance to a common standard for commoditization. Some of the projects representing each approach are: Open Compute Project (OCP), whose mission is “to apply the benefits of open source to hardware and rapidly increase the pace of innovation in, near and around the data center and beyond.” 9 OCPs Telecom Working Group has developed the CG-OpenRack-19 specification. This specification offers telecom data center operators the benefits of open platform standards combined with the needed carrier-grade and environmental enhancements required for edge computing, 10 which will be one of the most important building blocks for successful 5G deployments. Disaggregated Network Operating System (DANOS) is an open and flexible alternative to traditional networking operating systems. DANOS will support a network operating system framework that leverages existing open source resources and complementary platforms such as switches and white box routers (note: the project is expected to be available in 2019). P4 is an open-source initiative designed primarily to provide a declarative language for interacting with networking forwarding planes. P4 programs specify how a switch processes the packets. P4 controls silicon processor chips in network forwarding devices such as switches, routers and network interface cards. O-RAN alliance use two themes: “openness and intelligence” for the next generation wireless networks and beyond: “Building a more cost-effective, agile RAN requires openness. Open interfaces are essential to enable smaller vendors and operators to quickly introduce their own service” and “Intelligence Networks will become increasingly complex with the advent of 5G, densification and richer and more demanding applications. To tame this complexity, we cannot use traditional human intensive means of deploying, optimizing and operating a network. Instead, networks must be self-driving, they should be able to leverage new learning-based technologies to automate operational network functions and reduce OPEX” Leveraging open source is important for enabling a high-performance, flexible 5G user plane. There are various open-source networking initiativessuch as Data Plane Development Kit (DPDK), Vector Packet Processing (VPP), Fast Data Input/Output Project (FD.io), Mobile Central Office Re-architected as a Datacenter (M-CORD), National Ground Intelligence Center (NGIC) and Open Virtualized multilayer Switch (Open vSwitch or OVS) that provide the necessary optimizations, bringing in the ability of the user plane to scale and handle increased throughput necessary for 5G use cases and services. 9 opencompute/about/ocp-adoption. 10 15 The Status of Open Source for 5G 4.2 5G RADIO NETWORK 5G brings a diverse set of requirements and use cases, requiring an entirely new RAN architecture that is flexible, modular and supports open interfaces. The new RAN architecture needs to be operationally efficient and able to dynamically adapt to various, diverse requirements of 5G. As shown in Figure 3.3.2 in the 3.2 5G Radio Network section, the Cloud-RAN (C-RAN)/Fronthaul Architecture separated the CUs also known as Baseband Units (BBUs) and the DUs also known as Remote Radio Heads or Units (RRHs or RRUs), with the CUs located at Central Offices (CO) or master cell sites while DUs are located at cell sites. These separately located units are connected via Common Public Radio Interface (CPRI) or gs 3.x. Until now, the main fronthaul standard was CPRI, which is outside the purview of 3GPP. Its successor, evolved CPRI (eCPRI), is implemented in a proprietary, non-interoperable manner. Also, O-RAN created a complete spec on Open Fronthaul Interface. Additionally, the C-RAN architecture is no longer able to handle the ultra-high radio speed, ultra-low latency and massive connectivity requirements of 5G. As part of the new RAN architecture, various functional splits have been proposed, each offering trade-offs such as reduced fronthaul capacity and higher latency. The key to effectively implementing the new RAN architecture with flexible splits and efficient fronthaul is the openness in its specification and implementation. Such proprietary software and interfaces are often tied to the underlying hardware, which is a significant roadblock for openness. Enabling multi-vendor, best-of-breed flexibility in the RAN requires a move away from proprietary hardware to off-the-shelf, general-purpose processing platforms. Adoption of such commodity network hardware will require a reference design with standardized interfaces. There are several industry forums such as O-RAN, Open RAN (ORAN) / Telecom Infra Project (TIP), which are focusing on decoupling the RAN control plane from the user plane, building a modular RAN software stack that uses commodity hardware and publishing open north- and south-bound interfaces. The challenge has been the lack of openness in the RAN architecture, which is being addressed with the specifications being defined by the ORAN group. A RAN architecture with disaggregated software based on open specifications running on commodity hardware could allow operators to reduce complexity, innovate faster and significantly reduce deployment and operational costs. Hence adoption of this is most applicable for 5G RAN architecture. With well- established backing from the telecom industry, a growing community and an open hardware ecosystem in communities, such as the Open Compute Project (OCP) and O-RAN, the open specifications and implementations coming out of these forums are, over time, likely to see higher adoption and thus should be considered as applicable. 4.3 5G CORE NETWORK The core network is a critical component, so it needs to be robust, highly resilient and high performance. 3GPPs Service-Based Architecture (SBA), has standardized the Network Functions (NFs), their procedures and the inclusion of NF sub-modules. Standards also define the APIs to be used by providing data model, protocol and format. However, there still exists lots of innovation and ongoing research in the open source community and within vendors to address specific problem areas or new ways of implementing specific interfaces. The following sections identify some ongoing open source initiatives in the control and user planes of the 5G core network. 16 The Status of Open Source for 5G 4.3.1 CONTROL PLANE (CP) Unlike previous generations, the 3GPP 5G system architecture is using open APIs referred to as Services Based Interfaces (SBIs). The communication over the SBI will need a services framework implementing a message communication bus that can support an efficient mechanism for synchronous communication between the NFs. Different mechanisms are needed for asynchronous data movement, such as transferring event notification, performance data, service availability and discovery. As of now, this has been outside the purview of the standards. Web-scale companies have widely adopted microservices architecture; hence there is some open source software that provides a microservice message communication bus. Some of the open source implementation of a communication bus for synchronous communication are reverse proxies such as NGINX, High Availability Proxy (HAProxy) and Open Network Automation Platforms (ONAP) Micro-services Bus Project (MSB). Apache Kafka streams are also a possible implementation for asynchronous data movement enabling real-time data ingestion. This is important in the areas of analytics-based automation and service assurance, where real-time event monitoring becomes highly critical. The Network Repository Function (NRF) will allow every network function to discover the services offered by other NFs. Open source implementations of Consul and others provide similar service registry and discovery mechanisms, which have been widely deployed by web-scale companies. SBI and NEF have adopted the Open API Initiative, which defines a standard, language-agnostic interface to RESTful APIs. 3GPP ha