Higo Information and Tencent Cloud’s open-source Cube Sandbox project completed deep joint optimization, running the AI agent sandbox on Haiguang’s C86-series CPUs.
According to the official statement, this adaptation involved three layers: application architecture, microservice routing, and Linux kernel scheduling. It systematically addresses scheduling jitter in high-concurrency MicroVM scenarios, tail latency, and cross-NUMA memory access latency. In the governmental, finance, and energy industries, AI Agents have been localized for domestic deployment. For the first time, a hardware-level secure sandbox foundation is in place.
According to the official statement, this adaptation involved three layers: application architecture, microservice routing, and Linux kernel scheduling. It systematically addresses scheduling jitter in high-concurrency MicroVM scenarios, tail latency, and cross-NUMA memory access latency. In the governmental, finance, and energy industries, AI Agents have been localized for domestic deployment. For the first time, a hardware-level secure sandbox foundation is in place.