Evaluating Virtual Cluster Configuration and Compiler Measured

Abstract

It is expected to meet improving IO efficiency, lower management costs, higher storage utilization at the same time when creating a virtual cluster on a host machine. To meet these three goals, many existing methods improve the host storage utilization and improve virtual machine performance by increasing number of nodes and adopting COW technology and QCOW2 image format. However, adding more nodes to the link path results in long path. Usage of COW and QCOW2 cannot effectively reduce carrying costs on the physical machine. Zpools management systems (ZFS) which reduces redundant data greatly improve data search requirements time. However, this method results in occupation of CPU performance. Thus, how to reconcile and balance will be primary considerations for us. For the above reasons, we first test all virtualization cluster performance under different operating systems. Next, we use different compilers to test performance and then validate the testing results empirically. By changing the hard disk storage capacity and cache size, we then use different compilers to test the effectiveness in order to find the best configuration of virtual machines.