A press release crossed my desk on September 6, 2007 from Virtual Iron concerning some very strong benchmark results. Here's a segment of their release.
Virtual Iron Software (www.virtualiron.com), a provider of enterprise-class server virtualization and virtual infrastructure management software, today announced its support for the latest Quad-Core Intel® Xeon® Processor 7300 series platforms. With the launch of its latest release, Virtual Iron is among the very first virtualization software solutions to commercially support the new Quad-Core Intel Xeon processor 7300 series. This enables Virtual Iron users to gain improved virtualization performance with less overhead for the most demanding workloads on the new Intel platform.
Quad-Core Intel Xeon processor 7300 series bring the energy-efficient virtualization performance of Intel® Core Architecture to MP server platforms. This platform delivered up to 167% virtualization performance improvement on Virtual Iron Version 4.0 software based on the vConsolidate performance benchmark, a methodology to accurately and consistently measure performance in a virtualized environment. The results were remarkable. The Quad-Core Intel Xeon processor 7300 series-based server utilizing Virtual Iron 4.0 was able to run twice as many Consolidated Stack Units (CSUs) relative to Intel’s previous dual core MP platforms. This unprecedented scalability translates into increased virtualization performance, more VM capacity and higher consolidation ratios improving even further the benefits of virtualization for enterprise data centers.
While very impressive, what does this really mean for decision-makers? On the one hand, these results look very impressive. On the other, it's just another number that will result in confusion in the minds of decision-makers unless it is used properly. With no disrespect to the good folks at either Virtual Iron or Intel, I almost always find benchmark results suspect.
"As a general rule the most successful man in life is the man who has the best information." said Benjamin Disraeli. He also is supposed to have said "There are three kinds of lies: lies, damned lies, and statistics." I'm sure he would have added benchmarks onto the end of that list had he been aware of them.
In the end, no particular benchmark will tell an organization exactly what performance they'll see in their own environment. Only a carefully run pilot project will reveal that data. A properly chosen benchmark can at least make it clear whether a given configuration has a hope of supporting the organization's workload.
In this case, I would recommend that organizations evaluate Virtual Iron's product first hand, that is, see a demo and then try a pilot project. While it is impossible for me to know the results ahead of time, I would not be at all surprised that the results will be pleasing.
Has your organiztion used Virtual Iron? If so, what kind of results have you gotten?