2024年6月22日上午，在华为开发者大会2024 “Deep Rooted in Tech, Lead the Future” 海外高峰论坛上，国内外嘉宾和专家齐聚一堂，交流华为领先技术及优秀行业案例。 此间，华为云PaaS产品部海外产品总监肖振华在 “A Leading Global Automaker's Serverless Approach to Real-Time IoV Data Processing for Tens of Millions of Vehicles” 议题中分享了华为云FunctionGraph函数工作流在世界销量领先车企的实践故事，表示：
The world's top automaker used FunctionGraph to build a serverless IoT platform that delivers scalability, stability, speed, and concurrency, with a 50% improvement in resource utilization.
IDC predicts that by 2025, the global number of connected cars will be 78.3 million, and the global penetration rate of intelligent connected passenger cars has exceeded 50%. As the core system of car companies, the Internet of Vehicles has the characteristics of high concurrency vehicle access, and the access of cars has the characteristics of typical peaks and troughs, with a large amount of morning and evening peak access.
In order to meet the business requirements of 67 million connected vehicles in the next 10 years, the world's leading large car companies have built their own full-life cycle vehicle management platform, which should be able to carry minute-level vehicle data reporting, 100T data increment per day, and support at least 10PB or more of stock. It is a huge challenge to build a stable Internet of Vehicles system that supports tens of millions of vehicles, especially for car companies whose main capabilities are in R&D and manufacturing rather than IT, they will inevitably face: high operating costs, new requirements such as classified protection of laws and regulations, complex business, complex operation and maintenance, etc., through the analysis of business pain points and construction requirements, the car company found that the serverless system architecture will be a suitable, modern and advanced technology selection.
Based on the use cases and experience of the enterprise and other car companies (as well as Huawei's serverless product features), HUAWEI CLOUD provides a reference architecture for serverless Internet of Vehicles
Illustration of the serverless IoV reference architecture tailored by HUAWEI CLOUD for the car company
Scenario scenario
After receiving the data from the vehicle for processing, it will be stored in the big data data of HUAWEI CLOUD and the AI platform for further processing.
Schematic design
Depending on the terminal device, APIG or IoTDA is selected in the data access layer of the data. Regardless of whether it is APIG or IoTDA, the ingestion data can be directly transferred to FunctionGraph and Huawei Cloud's Function Compute service. Your business code is deployed in FunctionGraph, and after the data is triggered for execution, part of it is written to OBS or a database to complete the processing. You can also divide the entire processing process into multiple function programs, and the data processed by the first function is written to the message queue, and then the next function processing is triggered to decouple the business process. Eventually, the data falls into storage and other business systems.
Solution Advantages:
High elasticity, industry-leading minute-level 5000+ function instance elasticity: FunctionGraph provides faster scaling capabilities to enable users to run their business stably. The core capability of FunctionGraph is to quickly and reasonably elastic instances based on traffic. (1) Through the intelligent and hierarchical resource preheating covering the three layers of nodes, resource pools, and functions, the Internet of Vehicles business can obtain resource processing services faster, (2) through the self-developed SnapShot preloading capability, the cold start of the acceleration function, and (3) the instantaneous hierarchical scheduling algorithm can schedule traffic according to the cluster situation, so that the elastic capabilities of multiple clusters can be superimposed. In this practice, FunctionGraph can reach a speed of 5000+ instances per minute to scale the instance to a single user. Better support the peaks and troughs of tens of millions of vehicle access.
Illustration of "High Elasticity".
High stability: A more robust system, end-to-end 3w+QPS 0 error for the Internet of Vehicles service. The computing platform on which the Internet of Vehicles relies must be stable enough to ensure the stability of the upper-layer business, and customers can make billions of calls per day during operation, and can achieve zero errors in a single day. This is based on the customer's good engineering, which also requires a high degree of stability of the platform. At the FunctionGraph service level, the natural management and control of FunctionGraph function workflows and the business plane are designed with high availability of multiple AZs, and customer requests may be executed in multiple AZs and rescheduled based on system and user configurations after an exception occurs, ensuring high availability of cloud services and users. To achieve natural support for disaster recovery, one person can complete the operation and maintenance of tens of millions of Internet of Vehicles.
Illustration of "High Stability".
High concurrency & low latency: Faster real-time processing of 10GB/s data in seconds. To better support users' requirements for low-latency data processing, the HUAWEI CLOUD team helped customers select the single-instance multi-concurrency capability of FunctionGraph workflows. This feature allows a function instance to process multiple requests, increasing the request throughput per resource and further reducing the cold start latency. This allows more resources to be devoted to processing and consuming operations. Through the multi-instance consumption capability, you can reduce the accumulation of data in the process of function processing, so that the business side can feel the real acceleration. After the Internet of Vehicles (IoV) is migrated to FunctionGraph, the latency of the original platform changes from minutes to seconds, and the parallel execution also reduces the processing time of the function, which is more than 20 times lower than that when the single-instance multi-concurrency feature is not used.
High concurrency & low latency
In summary, serverless helps enterprises simplify development and deployment in the Internet of Vehicles scenario, and achieve rapid expansion and low cost. HUAWEI CLOUD FunctionGraph can help enterprises reduce TTM from weeks to days, greatly reduce costs, improve resource utilization by 50%, and implement millisecond-level capacity expansion to cope with traffic peaks, helping enterprises build service systems with high elasticity, stability, concurrency, and low latency.
Follow #HUAWEI CLOUD Developer Alliance#Click below to learn about HUAWEI CLOUD's fresh technologies for the first time~
HUAWEI CLOUD Blog_Big Data Blog_AI Blog_Cloud Computing Blog_Developer Center-HUAWEI CLOUD