A coflow scheduling method based on deep reinforcement learning

Abstract

The invention discloses a Coflow scheduling method based on deep reinforcement learning. Based on deep reinforcement learning and combined with multi-level priority queues and basic flow scheduling algorithms, the Coflow scheduling strategy is formally expressed by establishing a Markov decision model, and then Complete the adjustment of the Coflow scheduling strategy. The present invention uses the data size, the total number of flows, the number of active flows and the duration of Coflow in the current network environment to form a state matrix to represent the network state, define the action of adjusting the flow transmission multi-level priority queue, establish an evaluation equation, and realize the Coflow Minimal optimization for average completion time. The present invention combines deep reinforcement learning and multi-level priority queues, adjusts the Coflow scheduling strategy through a lot of trial and error and learning in the Coflow sending scheduling environment, and incorporates more environmental parameters compared with the traditional heuristic algorithm, effectively improving the Coflow transfer efficiency.

Description

technical field [0001] The present invention relates to Coflow scheduling, and more particularly, to a Coflow scheduling method based on deep reinforcement learning. Background technique [0002] Big data computing frameworks headed by Apache Spark are often used in computing tasks such as data mining and machine learning. They use a distributed architecture to divide the input data and assign the divided small data blocks to different computing nodes for operation. , and then redistribute and calculate after the data processing is completed. After one or more iterations, the desired result is finally obtained. With the execution of the above computing application, a large number of communication flows for transmitting data will be generated between different nodes (Nodes) in the cluster. These communication flows have high parallelism and occupy most of the application running process, which profoundly affects the computing process of big data applications. Therefore, fin...

AI Technical Summary

Problems solved by technology

The multi-level feedback queue has the advantages of small scheduling delay and can effectively distinguish between long flow and short flow, but there are still defects, that is, the threshold value often needs to be manually set in advance, which is empirical, and the network environment is complex and changes frequently. Thresholds are often not optimal, which limits the final effect of scheduling

Images