A modeling of network I/O efficiency in Xen virtualized clouds

Cloud computing derives from the concepts of distributed computing, parallel computing and grid computing originally. Today, more and more commercial organizations and academic institutes join in the area of cloud computing for safety and low spending reasons. With plenty positive advantages, system virtualization is becoming a prevalent technology to build underlying infrastructures in virtualized cloud environment. Among a number of state-of-the-art Virtual Machine Monitors (VMMs), Xen is definitely to be a conspicuous hypervisor based VMM. However, there still lacks of in- depth understanding of the efficiency of I/O communication among multiple Virtual Machines (VMs) in both academic and industry worlds, especially the burdensome network I/O works. In this paper, based on the server side and client side, two groups of equations are proposed to mathematical modeling the exchanged memory pages in Xen I/O channel. According to our carefully designed experiments, the proposed functions calculated the number of exchanged memory pages accurately, with an average error rate around 6%, even in multiple VMs scenario. Meanwhile, we illustrated and analyzed the effects of varying MTU (Maximum Transmission Unit) size respectively based on our experimental results and proposed equations. Afterwards, an effective improvement for optimizing network I/O performance is also proposed in this paper.