Alibaba's Self‑Developed White‑Box Switches: Scaling, Automation, and Ecosystem Evolution

Since 2018, Alibaba has been designing and deploying fully self‑developed network hardware, covering the entire Alibaba Cloud network and forming the foundation of its hyperscale data centers.

Super‑large‑scale access : White‑box switches, based on the Scale‑Out concept and CLOS architecture, provide a low‑cost, horizontally extensible solution that replaces traditional chassis switches. SDN opens the system, and the SONiC operating system, promoted by Microsoft and Alibaba, standardizes the open‑source switch OS.

Super‑large‑scale operation : Automation replaces manual, driver‑based network management. Rich monitoring, centralized data analysis, and automated fault detection, isolation, and recovery enable efficient operation of tens of thousands of switches.

Super‑large‑scale evolution : White‑box hardware allows rapid deployment of new chipsets and architectures, reducing dependence on commercial vendors and lowering total cost of ownership.

Self‑development roadmap : Starting with a 12.8 T single‑chip, 128 × 100 G design, Alibaba evaluated 200 G vs. 400 G options and chose a 200 G path to preserve cluster scale while driving the ecosystem toward 400 G readiness.

Full‑link automation : The network now supports automated architecture verification, large‑scale deployment, version upgrades, and intelligent fault handling, eliminating manual CLI configuration.

Second brain (BMC) : The Baseboard Management Controller offloads control tasks from the CPU, providing fault detection, analysis, and automatic recovery without human intervention.

Second lifeline (out‑of‑band network) : Alibaba has also self‑developed out‑of‑band switches and serial‑port servers, integrating them into a unified, automated management framework.

Ecosystem power – S³IP : The S³IP initiative standardizes driver interfaces (sysfs), platform testing (PIT), and the D4OS operating system, fostering a collaborative white‑box ecosystem with dozens of partners.

Port standardization – QSFP112 : To support future 400 G ports, Alibaba launched the QSFP112 MSA, preparing the industry for the next generation of high‑speed optical modules.

Conclusion and outlook : Alibaba’s white‑box switch experience underpins cost‑effective, automated cloud networking and sets the stage for “predictable networks” that meet the bandwidth and latency demands of emerging AI and large‑scale compute workloads.