The improve of load balancing strategy on Linux system based on EP
Qiao ShiJiao, Xu Hong, Yuan XingDe · 2011
Although the Linux2.6 kernel contains a O(1) scheduler, it does not include a certain algorithm to analyze the task's total execution time, which will directly result in certain shortcomings in the computing speed. In order to minimize the task scheduling time of completing load balance, this paper introduced the EP algorithm (Evolutionary Programming) applied in Scheduler. When the system needs to migration process to achieve the load balance, the improved scheduler analyzes the task of busy CPU queue as an entity and then finds out the most optimal distribution solution through the global search ability of EP algorithm, which will eventually achieve the shortest total execution time. The experiment results demonstrate that the enhanced algorithm is more advanced in computing speed.