Director of R&D
Tim Dong, Director of R&D at iTutorGroup, is responsible for the “TutorMeet+” - the realtime interactive cloud platform. He graduated from Jilin University and worked in Internet companies such as Tuniu.com and Hujiang.com etc. He is responsible for architecture research and development and technical management. He has a deep understanding of the architecture design, distribution system, container and micro-services of large-scale Internet systems, and is good at system optimization and solving various intractable diseases in the process of system evolution. In recent years, focused on the development of WebRTC technology. The languages used in the work are mainly Java, C#, Golang and so on.
Track: Service Port &Architecture Design
2019/04/19 14:00 - 17:55
Different from traditional internet applications, now internet applications based on audio and video has a new set of requirements for stability, real-time and extensibility of service port. We find that conventional solutions are no longer applied in special scenes, such as cache. We can’t build the house on sand when face challenges, and we must use more reasonable architecture design for suiting new system requirements. The key point of our topic is how to build new generation internet platform with high quality, high concurrency and high stability. Welcome to join us！
Topics in the track
Introduce the architecture roadmap of Bilibili uploading systen and experience of optimisation. Suitable for firends whith similar needs.
This session will share the principles in designing the architecture of a distributed SFU/MCU media server in Intel, and the design decisions in resolving the key issues in making a distributed SFU/MCU media server in real. Those design decisions include how to define the concept model regarding hierarchization and modularization, how media data is transmitted among components in the cluster and the MNCP (Media Node Control Protocol), and how the fault-tolerance mechanism works in such a distributed SFU/MCU. This session will also describe how MSOA (Micro-service Oriented Architecture) is achieved in Intel CS for WebRTC.
Nowadays, the live webcast is in full swing. The live broadcast of the VOD (Video On Demand) is one of its application branches. The anchor broadcasts while watching the VOD, and the viewer simultaneously watches the mixed video stream of the VOD and anchor comments. This sharing will introduce the video mixing scheme based on the IQIYI Hydra platform. The anchor will push the WebRTC video to the cloud, and the cloud will pull the corresponding VOD stream and mix them up, and then encode and push the mixed stream to CDN. I will focus on why I need to mix the streams on the server side, how to improve the success rate of WebRTC media transmission and the quality of audio and video, how to synchronize the progress of decoding of the VOD stream on the client and the cloud side, and how to achieve high concurrency and availability of the video mixing services.