A Ｒeinforcement Learning of Cloud Ｒesource Scheduling Algorithm Based on Adaptive Weight
We considered the cloud computing resource scheduling problem，and proposed a multi-objective optimization mathematical model to optimize task completion time and running cost simultaneously. To solve the model，a reinforcement learning based on adaptive weight algorithm was presented. To improve the...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | zho |
| Published: |
Harbin University of Science and Technology Publications
2021-04-01
|
| Series: | Journal of Harbin University of Science and Technology |
| Subjects: | |
| Online Access: | https://hlgxb.hrbust.edu.cn/#/digest?ArticleID=1937 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | We considered the cloud computing resource scheduling problem，and proposed a multi-objective optimization mathematical model to optimize task completion time and running cost simultaneously. To solve the model，a reinforcement learning based on adaptive weight algorithm was presented. To improve the efficiency of the algorithm，we designed the state space and the action representation method based on one-dimensional array to compress the storage cost of algorithm. Furthermore，an adaptive weight was introduced in the strategy of action selection to improve the convergence speed，and the weight was updated adaptively according to the return value after each execution of the action，then the action selection strategy was determined by using the heuristic function. Extensive simulation experiments were made with randomly generating data on Cloudsim platform. Ｒesults show that the heuristics accelerate Q-Learning based on adaptive weight is better than genetic algorithm and Q-Learning algorithm in optimization ability and load balancing. The convergence rate of the heuristics accelerate Q-Learning based on adaptive weight is faster than Q-Learning algorithm and heuristics accelerate Q-Learning algorithm. |
|---|---|
| ISSN: | 1007-2683 |