A Machine Learning Approach to Live Migration Modeling


Dynamic management of resources in data centers improves data center utilization in terms of energy consumption and device utilization. Virtual machines are among techniques to achieve this goal. To enable the VM to adapt to dynamic fluctuations of workloads, Live Migrated virtual machines are introduced, which stands for moving a VM from one physical host to another while the VM keeps running. Live migration enables load balancing, fault tolerance, ease maintenance, etc. However, live migration is costly in terms of data traffic. Having an understanding of performance metrics of different live migration algorithms is important so that one can rely on an efficient algrothm to migrate a running VM is important. Therefore, this work uses a Machine Learning (ML) approach to predict key metrics of different live migration algorithms, and then find the best suitable choice for live migration. Indeed, given resource usage as well as the characteristics of VM work load, this paper predicts six key metrics of live migration, such as: total VM migration time, total amount of data transferred, VM downtime, performance degradation of the VM, and CPU and memory usage on the physical hosts, for five different live migration algorithms.



This work uses machine learning to not only capture which algorithm to choose for live migration, but also targets to find which metrics should be considered while choosing an algorithm, and the main reason for such approach is that machine learning is an ideal technique to automatically generate models for the different metrics and the available live migration algorithms while using profiling data gathered in data centers.



Strengthes:



+ Considered approach can be implemented into existing migration frameworks to select the best live migration algorithm for the migration of a VM. Indeed, it answers which algorithm to select in a given situation. They have shown 2-5 time better performance can be obtained by such accurate live migration algorithm choosing.

+ Considered model can predict several target metrics for all commonly used live migration algorithms in a flexible and automated manner.

+ Prediction vales for metrics are very accurate.

+The idea enables efficient utilization of resources such as amount of time, network bandwidth, or host CPU and memory usage.



Weaknesses:

-         The considered model parameters seems more like heuristic, and there is not enough justifications for such options. One easily may consider many other different features to represent the model. 




Comments

Post a Comment

Popular posts from this blog

A Machine Learning Approach to Live Migration Modeling

StormDroid: A Streaminglized Machine Learning-Based System for Detecting Android Malware

Coordinated Management of Multiple Interacting Resources in Chip Multiprocessors: A Machine Learning Approach