For the needs of real-time parallel processing of critical task and real-time state control in muti-node computing system, a lightweight distributed processing computer mechanism is designed for iterative computing type tasks based on theoretical techniques such as memory computing and distributed computing. Using the feature of distributed collaborative processing, the data is deployed to different nodes to reduce the burden of each node and. To support continuous computational iterations, temporary data for each computation is temporarily stored in memory to reduce the disk IO access time during the iteration. Before the start of the whole task, the transformation relationship between the saved data is defined by generating the run plan graph of this task. In addition to load balancing of processing nodes, a balanced scheduling mechanism of control nodes is introduced to monitor the operation status of each node in real time through load balancing and control the deployment of computation tasks. Tests show that the designed lightweight computing mechanism outperforms the traditional distributed computing mechanism. In addition, the performance outperforms the mainstream memory-based distributed computing mechanism when carrying multiple tasks.