情報処理学会 第76回全国大会講演要旨
5X-9
An Agent-based Approach for Mining Performance Data Streams from Virtualized Environment
Virtualization technology is becoming popular because it facilitates faster deployment of services. However, there are still some issues such as effective resource utilization. In order to save energy, VM resources must be accurately predicted and timely provisioned to save resources and maintain service levels respectively.
We propose a knowledge based approach for recommending new VM instances in order to cope with resource utilization. Based on agent technology, the approach uses a recommender agent and streaming agents. A Recommender agent autonomously mines streams from multiple streaming agents using ensemble-based mining techniques such as hoeffding tree.
We propose a knowledge based approach for recommending new VM instances in order to cope with resource utilization. Based on agent technology, the approach uses a recommender agent and streaming agents. A Recommender agent autonomously mines streams from multiple streaming agents using ensemble-based mining techniques such as hoeffding tree.
