A global memory sharing method includes counting, by the sub-operating system, a page replacement rate in a task scheduling period of a predetermined quantity of times, and a memory residence time ratio in a page replacement period; calculating, by the sub-operating system, a memory pressure index according to the page replacement rate and the memory residence time ratio; and if the memory pressure index is greater than a memory pressure threshold, sending, by the sub-operating system, an application to a global memory management service module. According to the forgoing method, in an architecture of multiple operating systems, each sub-operating system can complete much adaptive work, which reduces complexity of the global memory management service module and improves system performance. Meanwhile, a utilization rate of global memory may be improved, and the architecture of the multiple operating systems may have better extensibility.