The blogger has not authorized any person or organization to reprint any original articles of the blogger, thank you for your support for the original!
blogger link
I work for an internationally renowned terminal manufacturer and am responsible for the research and development of modem chips.
In the early days of 5G, he was responsible for the development of the terminal data service layer and the core network. Currently, he is leading the research on technical standards for 6G computing power networks.
The content of the blog mainly revolves around:
5G/6G protocol explanation
computing power network explanation (cloud computing, edge computing, end computing)
advanced C language explanation
Rust language explanation
Article directory
1. Introduction to 6G data plane
Our company published an article about 6G data plane last year, which introduced in detail why 6G data plane is needed, data plane architecture and implementation details.
1.1 Why does 6G need a data plane
-
New capabilities such as communication perception and AI in 6G will generate massive amounts of data. These data can come from terminals, edges, gNBs, and CNs. The existing 5G user plane is not suitable for carrying these data. The specific reasons are given below:
5G user plane bearer 6G data plane data bearer Function PDU sessions provide end-to-end user plane connections between user equipment and network equipment Distributed data pipelines are composed of functions such as data collection, preprocessing, forwarding, storage, and analysis start and end point UE and UPF Any network element and terminal equipment data forwarding Forwarding devices only forward packets On-road computing needs to be implemented: in the data pipeline, data is transformed and optimized while being forwarded, so as to reach a state that can be analyzed and applied forwarding principle Packets are forwarded based on destination address Packets are forwarded based on data service and data pipe identities Topology point-to-point connection arbitrary topology -
Most of the data in existing 5G is point-to-point, while the data in 6G is in a distributed state (for example, perception data, AI data, network behavior and state data, etc., may come from different devices and finally converge to one device for calculation) , so any topology needs to be supported;
-
There are data islands in the 5G network. For example, there may be a problem of repeated data collection between gNB and CN, which means that it is difficult to achieve data sharing (not only within the mobile communication system, but also across domain manufacturers);
-
A lot of data in 5G is one-time and will hardly be stored, such as measurement information reported by terminals, etc., and this information can be used by endogenous AI in 6G to improve user experience and optimize network performance, etc., so 6G data Surfaces can provide a global mechanism to persist data.
-
Some of the existing data analysis functions of 5G are aimed at data in specific fields, and cannot be used to handle general data service management of 6G communication networks. A comparative analysis table is given below:
NUDAF MDT (Minimize Drive Test) SON (Self-Organizing Network) ETSI-ZSM ETSI-PDL IEEE-2144.1 deployment mode layered centralized Distributed, centralized or hybrid layered distributed centralized function, advantage Driving network automation and service orchestration through network data analytics in 5GC Network Planning and Optimization Network performance and user experience optimization Automate network and service management Federated Data Acquisition, Data Sharing, and Computing Feasible IoT management and data services data collection Aggregate data from data sources UE and RAN nodes collect data UE and RAN nodes collect data E2E Data Acquisition IoT device acquisition Trusted perception, data generation and collection data source 5GC NFs，OAM，AFs UEs EMS Infrastructure resources and network services Different Organizations Owning IoT Devices IoT devices open service by NEF not support not support via integrated bus not support via the data API Trusted (Security, Privacy) user permission required user permission required not support not support Based on distributed ledger Based on blockchain Example Network status (road congestion and early warning information, etc.) and device behavior (UE mobility, etc.) monitoring Coverage optimization, mobility optimization, capacity optimization, public channel parameterization, QoS verification Self-configuration, self-optimization, self-healing, self-protection, self-learning E2E network and service management; integration and interoperability; security and traceability smart city; smart healthcare; smart transportation All IoT applications Area 5GC RAN RAN OAM UE、AN、CN、TN IoT&Wireless Network standards organization 3GPP 3GPP 3GPP ETSI ETSI IEEE
2. A possible data plane architecture
The following figure is the data plane architecture figure 1 given in the Huawei paper :
2.1 Functional entities
-
Data Orchestrator (DO, Data Orchestrator): Responsible for coarse-grained, non-real-time data orchestration, it is a portal for receiving data service requests, which will convert data service requests into corresponding data pipeline construction requests and send them to DC. DO is also responsible for cooperating with other network services. For example, computing power network service arranges computing power, while DO arranges data. At the same time, a data security protection and privacy protection technology library (DPTR) is built in DO, including differential privacy, homomorphic encryption, zero-knowledge proof, etc., providing data security and privacy protection capabilities, and on-demand data protection technology (DPT ) is empowered to DA.
-
Data Controller (DC, Data Control): Responsible for fine-grained real-time orchestration tasks, combining data pipelines in the local domain according to DA capabilities and data service requests. The collaboration of DO and DC can realize the flexibility and programmability of data pipelines. Secondly, the DC will receive the DA's capability report and implement the registration and revocation functions of the DA, and realize the real-time supervision of the DA by detecting the heartbeat of the DA.
-
Data Agent (DA, Data Agent): Execute other services such as data acquisition, data preprocessing, data storage, data analysis, and data sharing orchestrated in the data pipeline. The data store among them is responsible for the local storage of a small amount of, or short-term, or data with privacy protection requirements. Can be built into network functions or deployed separately. The DA reports the data service capability to the DC, and then the DO/DC selects the appropriate DA based on the service request and the DA capability, and implements the arrangement. The DO will update the DPT on the DA as needed. DA provides data services externally through the service API.
-
Data Storage Function (DSF, Data Storage Function): It is mainly used as a storage expansion component of DA for large-scale data storage or long-term storage.
-
Trusted Anchor Agent (TAA, Trust Anchor Agent): An independent component defined in the data plane architecture to ensure the credibility of 6G data. Primarily responsible for protecting data confidentiality, integrity and reliability.
The following figure is the detailed architecture figure 1 given in Huawei's paper :
2.2 Operation of DO and DA
2.2.1 Operation of DO
The DO obtains the global information of the DA network through the data service capability reported by the DA and the logical connection status between the DAs. Then, DO selects the appropriate DA according to the received data service request, arranges the data pipeline, and calculates and constructs the data forwarding path. DO sends data forwarding information to DA through Data Forwarding Control Protocol (DFCP), and updates and deletes data forwarding information as needed. Finally, DO collects statistical data from DA through DFCP.
2.2.2 Operation of DA
DA can realize various data processing functions, and these functions can be reported to DO/DC as DA capabilities during DA registration, and can report capability updates in time. DO implements orchestration according to the specific service requirements of the application and the data function of the DA. The tasks initiated by the business logic of the network function (NF) enter the DA, and the DO orchestration strategy executes the specific processing functions of the DA. If the cooperative work of multiple DAs is required, after the last function of the local DA is executed, it is transferred to the next DA through the data forwarding function, or provided to the application 1 through the data service API after the local DA runs all functions .
2.2.3 DA deployment
As shown in Figure 1 below , DA has three feasible deployment modes in the network: independent, built-in, and hybrid.
- Independent: that is, DA is deployed in the form of an independent network element or NF in the network;
- Built-in: Indicates built-in DA in RAN or NF and terminal equipment;
- Hybrid: Indicates that an independent DA is added, and a built-in DA is deployed in the RAN, NF or terminal equipment at the same time;
2.3 Data Forwarding Technology
In the 6G data plane, data management and processing adopt the form of pipelines. Data is not only transmitted through pipelines, but also completes functions such as collection, processing, storage, and analysis at the nodes that flow through. In addition, the establishment of a session in a traditional network is based on the construction of a communication path. Nodes (usually routers or switches) on the path are only responsible for forwarding session packets and not processing packets. In the 6G network, in order to meet the needs of new services and new scenarios, the nodes (DA) on the data pipeline need to process the packets as needed, and then forward them to the next node. Therefore, the 6G data plane needs to build a new data-oriented forwarding mechanism.
Sessions in traditional networks are established point-to-point, aiming to find a suitable communication path in complex network topologies. The 6G data plane is distributed, so the data pipeline (convergence and distribution of incoming data) needs to support any topology. Figure 1 below shows three 6G data plane data forwarding technical solutions.
2.3.1 DA is stateful, message is stateless
The data forwarding control entity (DO) arranges and composes the data pipeline and its topology according to the functions of the DA according to the data business requirements, and writes the data forwarding entry into the data forwarding table of the corresponding DA. DA forwards the data to the next hop according to the entry until the end of the forwarding entry. At the same time, the DA counts the number of data packets and bytes forwarded, and reports it to the DO as needed. After the execution of the data service is completed, the data pipeline is deleted, and the DA deletes the data forwarding entry.
2.3.2 DA stateless, message stateful
According to business requirements, DO arranges and composes the data pipeline and its topology according to the capabilities/functions of DA, and sends the data forwarding entries to the ingress DA. The ingress DA forwards the forwarding information as the header information of the data packet to the next hop. The DA in the forwarding path forwards according to the forwarding information carried in the header of the data packet, and deletes the forwarding information related to the DA. The egress DA deletes the address/identification information in the message header and submits it to the upper-layer application. The DA counts the number of data packets and bytes forwarded, and reports it to the DO as needed. The edge DA deletes the data forwarding entry for a given data service after the data service ends.
2.3.3 DA and messages are stateless
According to business requirements, DO organizes data pipelines and their topology according to DA capabilities/functions. The DO encodes the data forwarding path corresponding to the data service, and sends the encoding to the ingress DA. DA calculates the next hop of the data packet through the decoding operation, and forwards it to the next node after completing the data processing. The egress DA submits the message to the upper layer application. DA reports to DO as needed according to the statistical data carried in the packet. The edge DA deletes the data pipeline after the data service ends.
3. Summary
The above are some designs and ideas for the 6G data plane in Huawei’s paper. In fact, the discussion on the data plane has started in 2021. Huawei is also the first company to have papers and demonstration demos. In fact, Intel has also published an article on 6G cloud computing. The paper on the native system also involves the conception of the data plane, and the writing is also very in-depth. After that, I will sort it out and write a blog post.
