An Application Deployment Approach Based on Hybrid Cloud

摘要

Hybrid Cloud is a composition of private and publicclouds, which can assist an application to utilize the advantages of distinct clouds and achieve high availability, scalability and robustness. Despite its benefits, hybrid cloud can present special challenges such as cloud heterogeneity, inconsistent interface, communication overhead between clouds and etc. In this paper, we propose a novel prediction model, to automatically adjust the scale of applications according to real time resource monitoring, for the concern on scalability issues of applications in hybrid cloud. To evaluate the effectiveness and performance of ourapproach, we implement a prototype of hybrid cloud architecture which is capable of resource monitoring, auto scaling, virtual-machine-level load balancing and unified management. We conduct experiments to verify performance cost of Collectd, expanding and recycling virtual machines. Experimental results show that applications can achieve high scalability with atolerable overall overhead provided by resource monitoring, auto scaling and load balancing.