I'm always interested when new virtualization platforms hit the market, so when Oracle announced Oracle VM, I listened. After all, to me, databases have always been one of those "don't ask, don't tell" stories in the virtualization world, or quite simply, one of those enterprise applications that you probably don't want to virtualize in a production environment. While the announcement gave me a lot of informatio I’m always interested when new virtualization platforms hit the market, so when Oracle announced Oracle VM, I listened. After all, to me, databases have always been one of those “don’t ask, don’t tell” stories in the virtualization world, or quite simply, one of those enterprise applications that you probably don’t want to virtualize in a production environment. While the announcement gave me a lot of information about what Oracle was doing, it also left me asking questions. I was confused about performance claims, support issues and licensing. One thing that really grabbed my attention was the company’s claim that Oracle VM was three times more efficient than other server virtualization products. I’m very interested in improving virtualization performance – I’ve watched it blossom over the last seven years, so anything to continue that trend is perfect in my book. Unfortunately, the claim really stopped there as it didn’t provide any specific benchmark test plans or results. Brian Byun, VP of Global Partners and Solutions at VMware told me that his company really couldn’t comment on Oracle’s performance claim since Oracle hasn’t published any information such as performance metrics or specific software versions or vendors used in their benchmarking.However, VMware quickly responded in any number of ways, one of which was an interesting post on the company’s performance blog site, VROOM!, titled Ten Reasons Why Oracle Databases Run Best on VMware.Byun said the blog post contained results from a real-world customer performance test that shows how databases, specifically Oracle databases, run very well and are optimized on VMware ESX Server 3.5. The results show near-native transactional throughput (more than 90%) running Oracle databases, 50x the I/O headroom on a single VMware ESX Server 3.5 host compared to typical database I/O workloads we have observed on physical 4-processor systems, and linear scaling of Oracle DBMS workloads across virtualized SMP and across large systems running multiple separate database instances on the same machine. As a final comment on our discussion about Oracle’s performance claim, he added, “We look forward to evaluating Oracle’s reproducible performance results once they are available.”In a more subtle answer to Oracle’s virtualization announcement, VMware also recently published a new whitepaper, SQL Server Performance in a VMware Infrastructure 3 Environment, where VMware attempts to show how well Oracle’s competitor database product functions within a VMware environment. The whitepaper says, “Current data indicates that database applications running on individual physical servers represent a large consolidation opportunity. Over 50 percent of these database applications run on two-way symmetric multiprocessor (SMP) machines, and in 90 percent of the cases transaction rates are under 20 transactions per second. CPU utilization averages less than 10 percent and approaches 20 percent at peak levels. As might be expected, I/O and data transfer rates are also low. Not surprisingly, many such database applications have been successfully migrated to virtual machines running on VMware ESX Server systems…” The paper describes transaction processing workload performance in virtual machines using Microsoft SQL Server 2005 and VMware Infrastructure 3. And the paper’s primary goal is to prove that Microsoft SQL Server 2005 can successfully handle enterprise-level transaction-processing workloads when running inside VMware virtual machines. And without simply stating that it runs well or better than a competitor product, VMware gives specific test plans, configurations and performance results to prove a point. Oracle’s announcement also said that its database software would only be supported in an Oracle VM virtualization environment. So I asked VMware how this announcement would play out with current and future VMware customers who want to virtualize an Oracle database on VMware’s platforms. Byun said, “Oracle’s support policy, which hundreds of VMware customers use today to deploy Oracle applications and databases from test and dev to production, remains unchanged. Oracle will still handle support calls from customers for known problems in Oracle products that are virtualized. If an issue arises that is suspected to be virtualization-related, Oracle would refer the customer to VMware.”And when asked about Oracle’s position on its application licensing policy, Byun said, “customers tell us they want to deploy virtualization in a manner that is neutral to their application software licensing agreements. Today, customers are telling us that Oracle still needs to provide improved ‘virtualization-neutral’ licensing guidelines that will let them reap the benefits of virtualization, as many other leading ISVs have done.” He added, “We believe that with the introduction of Oracle’s Xen stack, customers will be in a better position to ask for that clarification. We plan to continue to work with Oracle and other ISVs to help move the industry toward licensing policies that are value- and revenue-neutral to customers and the ISV.” Software Development