Introduction
| Test
methodology | Scorecard | Slideshow
Six-part
series on virtualization | Performance
results | Test
archive
On the Mac side, we used a MacBook laptop with a 2.2GHz Core 2 Duo. It had 144MB virtual RAM (integrated Intel Graphics chip) shared with 3GB of dynamic RAM. We also checked graphics and features on a MacBook Pro. Each guest operating system and virtual machine was assigned 768MB of RAM and 128MB of VRAM where available (VMware's Fusion automatically assigned 128MB of VRAM, although there didn't seem to be an option to change the amount).
For the Windows XP/Vista host testing, we used an HP EliteBook 2530p with 3GB RAM, a 1.86GHz Intel Core2 Duo and an older HP Pavillion DV9000 (2.33Ghz Core Duo, 4GB DRAM, internall 200GB hard disk). We assigned Windows XP and Ubuntu guests a pre-allocated 8GB and 6GB disk space, respectively. We performed multiple tests with one virtual CPUs. Because Virtual Box didn't support multiple vCPUs, we didn't test them.
Parallels supports as many as four vCPUs (eight experimentally) and Fusion supports as many as two (four experimentally). We could not set more CPUs than our computer had. Also, our SPECjbb2005 benchmark had a time drift problem with Parallels. The problem arose from an option in Parallels that syncs host time with Mac OS X time. Turning that off solved the problem, and our results reflect accurate timing.
We tested the environment with just the VM running, meaning no additional applications were loaded (besides the Finder and other background services.)
Read more about data center in Network World's Data Center section.