Integrating cloud application autoscaling with dynamic VM allocation

As the popularity of cloud computing continues to rise, more and more applications are being deployed in public clouds. To conserve provisioning costs while achieving performance objectives, clients should automatically scale up and down applications deployed in the cloud to match changing workload demands. The cloud provider, on the other hand, should attempt to consolidate load onto highly utilized physical machines, in order to reduce wasted power consumption. We propose a new algorithm combining both the automatic scaling of applications with dynamic allocation of virtual machines, in order to meet the goals of both the cloud client and provider. We evaluate this algorithm against running separate, independent autoscaling and dynamic allocation algorithms and show that the integrated algorithm can achieve better application performance with a significant reduction in virtual machine live migrations.