基于聚类的PaaS平台流量监控的迁移研究

Transfer learning of PaaS platform traffic monitoring via clustering

为了满足Web应用的快速部署,自动维护和自动扩容的需求,从而产生了PaaS平台。但随之而来的问题是如何实时监控PaaS的流量。为了能够实现流量的实现监控,研究人员提出了利用聚类算法来实现自动分类,但数据在传送很容易受到外界因素的影响,从而导致采集的流量是失真的,因此根据这样的数据来聚类分析后的结果是不准确的。针对此问题,以模糊C均值算法为基础,借鉴知识利用的思想,提出了一种具有迁移学习能力的聚类算法。并将其应用到PaaS平台的流量实现监控中,从而能够快速识别流量,从而能够从极大的保证系统的稳定安全的运行。

The PaaS platform isstructured to realize the Web’srapid deploymentand to satisfythe need of Web’s maintainand dilatationautomatically.But,there is a urgent problem that how to monitor the flow of PaaSplatformat any time.In order to be able to implement traffic monitoring related researchers using theclustering algorithm is presented to realize automatic classifi cation,but the data in transmission is easilyaffected by external factors,which leads to acquisition of fl ow is distorted,so according to the data tothe results of cluster analysis is not accurate.To solve this problems,a new clusteralgorithm,based onFCM algorithmand transfer learning thought,is introduced.This new cluster algorithm is used to PaaSplatform and tomonitor the fl ow of PaaS platform at any time,so that can recognition fl ow quickly andcan make platform run softly and stably.