Short-flow real-time optimization method and system, and network transmission terminal

Abstract

The invention belongs to the technical field of cloud data center network transmission flow scheduling, discloses a short-flow real-time optimization method and system, and a network transmission terminal. The method comprises the steps that: the reinforcement learning architecture of the data center short flow real-time performance optimization method is constructed; a data center network flow characteristic acquisition module is used for monitoring flow characteristics in a data center and generating flow size distribution; a multi-level queue degradation threshold value is initialized, andthe multi-level queue degradation threshold value is calculated according to the internal flow distribution characteristics of the data center; and the threshold dynamic adjustment is used for dynamically adjusting the threshold size in a fine-grained manner to adapt to the flow size change of the data center so as to realize data packet division priority transmission. According to the invention,the threshold value based on reinforcement learning is adaptively matched with the flow and transmitted, the interactive real-time short flow transmission delay is reduced, the lagged transmission oflong flow with high bandwidth requirement in the queue is improved, and the real-time performance of network transmission is improved.

Description

technical field [0001] The invention belongs to the technical field of cloud data center network transmission flow scheduling, and in particular relates to a short flow real-time optimization method, system, and network transmission terminal. Background technique [0002] At present, data center network traffic transmission optimization is one of the core research issues of cloud environment data centers, and it is a focus of attention and research in the industry and academia. Due to the current differentiation and dynamic characteristics of network traffic in data centers, data The performance optimization of central network traffic transmission has become a very challenging problem. The research on network traffic transmission has always aimed at reducing the traffic completion time and improving the real-time performance of transmission. Cloud applications such as web search, social network and service recommendation all exist in the data center in the form of short traf...

AI Technical Summary

Problems solved by technology

The above existing technologies, due to the characteristics of dynamic changes and suddenness of data center traffic, there is a mismatch between the degraded thresholds of queues at all levels and the size of the traffic, and the optimization period is long, so they cannot automatically adapt to traffic changes. When it does not match the size of the traffic, the priority transmission efficiency of short traffic is low, and there is still the problem that short traffic is delayed by long traffic

[0004] Through the above analysis, the problems and defects of the existing technology are: there is a mismatch between the degraded thresholds of the queues at all levels and the traffic size in the existing technology, and the optimization period is long, so it cannot automatically adapt to the traffic changes. When the size does not match, the transmission efficiency of short traffic is low, and there is still the problem that short traffic is delayed by long traffic

[0005] The difficulty in solving the above problems and defects is: how to fine-grainedly adjust the demotion threshold of multi-level queues to adapt to the dynamic changes in data center traffic volume, and improve the real-time performance of short traffic transmission needs to be solved

Images