ByteDance

Background

ByteDance is a Chinese multinational Internee technology company founded in March 2012. It owns several well-known products, including the media aggregation service Jinri Toutiao, the short video app TikTok (known as Douyin in China), as well as Xigua Video and the social platform Lemon8, among others. ByteDance has also developed a range of professional software that incorporates AI technology, such as CapCut and the office suite Lark (known as Feishu in China). In addition, ByteDance also provides public cloud services dubbed Volcano Engine. In 2023, ByteDance launched the Volcano Ark, bringing in a number of third-party large model providers to create a MaaS platform. Volcano Ark is targeted to be the industry large model selection store. It leverages the capabilities of the Volcano Engine to provide fine-tuning, inference, and evaluation of models. Model users can browse through the store to choose from a variety of industry-leading large model platforms.

Challenges

ByteDance's data center networks face challenges and demands across several key areas.

• Firstly, network deployment and management of a vast number of devices are complex. Meanwhile, ByteDance requires real-time monitoring and fault prediction in diverse service scenarios to guarantee smooth operation and high security of all types of services. To meet high-speed transmission and rapid expansion requirements of data centers, ByteDance has an urgent demand for switches with high performance, scalability, and quick delivery and deployment capabilities. 
• In addition, effective security monitoring and management tools are essential for building a stable and reliable network. This network is crucial for ensuring service continuity and data security of ByteDance's data centers.

Contending with Challenges

Ruijie Networks has provided ByteDance with a comprehensive solution to address their high-performance computing and large-scale data transmission needs. Ruijie applies a flexible and controllable RoCEv2 solution on the RDMA network to guarantee high network transmission efficiency and stability for ByteDance's data centers. To enhance device monitoring and fault pre-warning capabilities, Ruijie uses the gRPC technology for MMU monitoring on switches, achieving millisecond-level monitoring data collection.

To ensure network security, Ruijie has deployed dual stacks throughout the network. This not only optimizes ACL and routing storage space but also safeguards security management of diversified services, guaranteeing a secure and reliable operation environment for ByteDance's data centers. In terms of hardware delivery, the number of Ruijie 100G single-chip switches delivered to ByteDance exceeds half of that of switches used in ByteDance's IDCs applying the multi-stage Clos architecture.

Customer Benefits

• Based on the 100G/400G DCN practice, Ruijie helps ByteDance quickly upgrade to 400G AIGC intelligent computing networks. In this ToB scenario, Ruijie has implemented flexible solution delivery, ensuring secure isolation of services among multiple customers and achieving coverage across various scenarios.
• Moreover, Ruijie has successfully enabled large-scale application of AI high-performance computing clusters. By providing high-performance network devices and solutions, Ruijie ensures efficient and stable running of AI computing clusters, meeting ByteDance's growing needs in the AI field. In addition, Ruijie achieves precise monitoring and real-time data collection of network devices, making monitoring data more reliable and reflective of network devices' status and performance. This ensures continuity and stability of services.
• To satisfy the changing and growing business needs of ByteDance, Ruijie is actively promoting the transition to the 100G/400G architecture and cooperating with ByteDance to build 800G networks. The 800G architecture will provide higher bandwidth and lower latency, enabling rapid response to ByteDance's business needs and ensuring efficient network operation. Ruijie's rapid iteration capability and continuous adoption of cutting-edge technologies will help ByteDance maintain a competitive edge and stay unbeatable in the market.