I still contend that published and documented benchmark reports with absolute scores are what matters. VMware has published ours as above and Oracle has not.

Eric

Your link is broken, but I typically don’t find those reports very usefull so it doesn’t matter. I believe the Tolly Group report is what Oracle bases their claim on but I’m more concerned with real world experience.

Just today an AS3AP insert test I ran with 50 users showed 2806,09 tps on Oracle VM 2.2 and 1812,52 on ESXi 3.5.

Granted this is out of the box performance on lab hardware, and the differences tend to be a little less when you involve production hardware and some real storage, but it illustrates what I experience myself and hear from others.

It’s almost certainly possible to get better perfomance from both solutions, but most customers seem unwilling to spend much time fiddeling with a low cost platform.

Could you provide some references to test results showing that Oracle VM “far outperforms” VMware ESX with Oracle databaseses? VMware has published test results of vSphere 4 running Oracle using a TPC-C based workload (http://www.vmware.com/pdf/Perf_ESX40_Oracle-TPC-C-eval.pdf). Those results show vSphere VMs delivering close to native performance with results scaling all the way up to 8-way VMs. Our report provides figures for DB transactions per second and IOPS in vSphere VMs that are far in excess of requirements for typical Oracle DBs.

I’m aware of the Tolly Group study Oracle commissioned that shows only Oracle VM results compared to native (no absolute TPS or IOPS figures) and only 2-way VMs. There’s nothing there backing up claims that Oracle VM outperforms vSphere.

Has Oracle published other test results?

It might be worth mentioning that Oracle VM, whatever shortcomings it otherwise has, far outperforms VMware for Oracle databases.

Oracle will support customers running its products on a VMware platform, but with some restrictions. If you have access to the Oracle support site, take a look at Metalink article 249212.1 for the details. In short, it says that Oracle does not certify its products in a VMware environment, but they will provide support, “for issues that either are known to occur on the native OS, or can be demonstrated not to be as a result of running on VMware.”

I’m not aware of any Oracle customers who have been denied support by Oracle because they were running on VMware. However, Oracle sales reps will frequently (and usually, unsuccessfully) bring up their lack of certification for VMware as FUD to drive customers to their own Oracle VM Xen hypervisor.

Eric

Being based on Xen, I actually find Oracle VM to be far superior to VMware ESX as a solution, simply because of the back-engine of Xen and the opportunities afforded by distinction being paravirtualization and hypervisor-based virtualization.

It must be noted that, with regard to TCO, one can reliably run a MySQL DB, 3 Apache 2 webservers, 2 JVM servers, a Perforce server, a Subversion server, and a bastion host on a single Xen server which costs $300 to build from parts from Fry’s.

That kind of functionality and flexibility, when implemented on enterprise-class hardware affords great headroom and from this flexibility comes the value proposition.

Let us bear in mind that Amazon S3/EC2 is based on Xen. And so is Oracle VM.

Therefore, I think it more fair to say that Oracle VM is a poor replacement for the user-friendly richness of VMware.

But as a solution stack, administered by intermediate to senior-level Linux/Unix/Solaris administrators; it can be tailored and configured to suit an environment and deliver great value.

Again, the idea of Xen (and therefore Oracle VM) is that the solution does not subsume the rest of the stack. Shadow-imaging is a back-end technology, so let the SAN do it. This is not Xen’s job or forte. Bonding and bridging is done at the OS level, therefore Xen doesn’t do it for you. LACP or PaGP models are configurable by the installer, there is no VLAN trunking interface on the web gui. Fault-tolerance can be designed via various servers in a server pool.

VMware is friendlier. But I believe Xen is more flexible and dynamic. The burden is upon the system architect to design how the solution is implemented. Which, I think, is how it should be.

The learning curve is far steeper, however.

Could you ellaborate on the following line as mentioned in your article :- ‘Oracle VM pools are only server pools and not resource pools. This prevents memory and CPU resources from being better managed and shared by VMs in the pool. You can’t carve out resource pools for your business and let them operate autonomously as you can with VMware.’