Industrial research firms such as Gartner and IDC are predicting an explosion in the number of online services in the coming years. Virtualization technologies could play an important role in such a world, as they create an opportunity to provide services in a cost-effective manner. However, to achieve ideal savings, more dynamic environments must be created, with Virtual Machines (VMs) being provisioned and altered on-the-fly. Management issues arise when using these elastic resources at scale. In this study, we provide an initial investigation of performance and scalability issues for image management in a virtualized data center. Results provided show that the choice of storage solution and access protocol matters. For example, our tests show the time to start a VM from a local hard drive under I/O intensive workload increases by a factor of 15 and for certain shared storage options, this factor increases to 30 times. Keywords virtualization, image management, performance, scalabil...
Tingxi Tan, Rob Simmonds, Bradley Arlt, Martin F.