Flow size estimation method and device

Abstract

The present application discloses a method and device for estimating traffic size, wherein the method includes: acquiring characteristic attributes of multiple historical flow sequences, and preprocessing the multiple historical flow sequences; Process the results and calculate the similarity between multiple historical flow sequences; cluster multiple historical flow sequences according to the similarity between multiple historical flow sequences to obtain multiple historical flow sequence clusters; according to multiple Clustering of historical flow sequences, creating multiple flow size estimation models; combining multiple historical flow sequence clusters and the similarity of new flow sequences, assigning flow size estimation models to estimate the flow size of new flow sequences, this method Clustering is based on the similarity of historical flows, and for each cluster cluster combined with the similarity of new flow sequences, a flow size estimation model is assigned to new flow sequences for flow size estimation, which improves the accuracy of flow size estimation , and can be deployed on a large scale.

Description

technical field [0001] The present application relates to the technical field of data processing, and in particular to a flow size estimation method and device. Background technique [0002] At present, in order to meet user needs, many large-scale services must respond quickly to user operations, for example, web search, social networking, retail, recommendation systems, etc. Therefore, the data center network corresponding to large-scale services should minimize the average traffic completion time to Meet user needs. In the related technology, the traffic size is estimated based on the historical flow ensemble learning model to reduce the average traffic completion time, that is, the traffic size is estimated by monitoring the statistical data in the network and learning from the historical flow sequence, so as to reduce the average traffic completion time. [0003] However, in the above techniques, in order to make the model more adaptable, it is necessary to input histo...

AI Technical Summary

Problems solved by technology

[0003] However, in the above techniques, in order to make the model more adaptable, it is necessary to input historical flows with large changes for learning, but the input difference exceeds a certain range, which may make it difficult for the model to effectively estimate the size of the flow. In addition, due to different There are differences between the flow patterns generated by applications. The learning-based mechanism needs to train a specific traffic size estimation model for each application. Once the parameters are changed or a new application is generated, it needs to be retrained, which is difficult to apply in practice on a large scale.

Images