Specifically, the computer and chip giants said they are working together on new tools to help IT select, deploy and measure virtualized server solutions for enterprise data centers.
One of the first tools to emerge from this joint initiative is a new virtualization benchmarking methodology called vConsolidate. The benchmark runs multiple instances of consolidated database, mail, Web and Java workloads on Intel-based IBM System x servers in multiple virtual CPU partitions.
She added that IT buyers could use vConsolidate to benchmark trial hardware or also to measure the potential benefits of server consolidation projects, a popular use of virtualization.
"As mid-sized and large enterprise IT organizations strive to cash in on the cost savings of data center consolidation through server virtualization, hosting applications on larger, more expandable multi-processor servers delivers the best return on investment," said Jim Northington, vice president, System x, IBM. "Nevertheless, many organizations need the tools to help them select the server platform that works best in their unique environments."
IBM and Intel have already used vConsolidate to measure IBM multiprocessor systems against competitors. The results showed an IBM System x3950 delivering up to 46 percent more performance than a competing system when running a mix of larger two and four virtualized processor partitions.
Based on these and other customer results, IBM and Intel also developed what they call a VMware Infrastructure Sizing Guide to help customers select and appropriately configure various virtualized server options. The sizing guide draws on IBM's long experience with virtualization technologies and techniques dating back to its early mainframe work.
Experimentation and results from vConsolidate and the sizing guide identify memory as a key limiting factor in determining how many virtual machines can be loaded on an Intel-based server.
Researchers said they've collected data on some 10,000 servers using an IBM consolidation and analysis tool. A key finding was that while virtualization increases total processor utilization, additional reserve memory is required for application usage spikes.