Use Case

13:30-14:15 UTC+8 , January 15

The Evolution of Trillion-level Real-time Messaging System in BIGO

BIGO currently has two major video products and services, Live and Likee. At present, BIGO Live's live broadcast business has covered more than 150 countries and regions and Likee short video also has more than 100 million users. Being well received and approved among young people, our products are popular all around the world.
In the past technical architecture, BIGO adopted open-source Kafka as a basic service to support real-time data processing and analyzing. And based on this, BIGO has built a complete recommendation infrastructure to provide users with high-quality recommendation services. However, as the business continues to develop rapidly, the scale of our processing messages has entered a trillion scale and the past architecture has encountered huge challenges.
Regarding these problems, we got in touch with Apache Pulsar. With the gradual deepening of Pulsar, we have benefited a lot from these excellent features brought by Apache Pulsar such as hierarchical persistent storage, low E2E latency and horizontal scalability. These advantages have also helped us overcome many difficulties in the production system.
This talk will introduce the development and construction of BIGO's messaging platform based on Pulsar. We will start with message processing, storage, etc., and deeply analyze the performance bottlenecks that may be encountered in the production environment, and share our production processing practice experience.

Speaker

Zhanpeng Wu

Apache Pulsar Contributor, Senior Engineer @ BIGO

Apache Pulsar Contributor, KoP Maintainer