AutoMQ for Automotive

XPENG Motors uses Apache Kafka to solve the log collection, processing, and analysis of various application systems on the cloud platform. The log data of various business applications is delivered to Kafka through a unified collection channel, and then Kafka distributes it to downstream basic components for consumption and processing.

Currently, this link supports multiple core scenarios and systems such as online monitoring and alarm, log retrieval, business operation data analysis, and security audit compliance. With the continuous growth of the business, two serious problems are gradually exposed:

• High resource costs: Using Kafka in the cloud, as the business scale grows, the cloud bill for the cluster gradually inflates.
• Heavy operational burden for scaling up and down: With the rapid growth of the business, there will often be a need for scaling up. Frequent operational changes burden Kafka, requiring careful attention to partition migration and traffic rebalancing.

AutoMQ completely decouples Kafka's computation and storage. It uses object storage like S3 to provide pay-as-you-go unlimited storage capacity. The stateless Broker in the computation layer allows the computation nodes to scale up and down rapidly, avoiding the cost increase caused by the waste of over-provisioned resources.

Through technical innovations, AutoMQ has successfully resolved the problems of our Kafka cloud bill getting out of control and the complexity of scaling up and down in the past.

Great Wall Motors has an extremely high demand for service stability, requiring that even if a fault occurs with one cloud service provider, it can immediately switch to another to continue providing reliable and stable services. Kafka, as a crucial component in modern data infrastructure, is extensively used in Great Wall Motors. Supporting this hybrid cloud disaster recovery, a Kafka solution that supports multi-cloud federation cluster deployment is required. At the same time, this Kafka solution needs to offer better cost efficiency and performance than Apache Kafka.