Oracle's database performs better under VMware than in a more expensive Real Application Cluster (RAC).
Oracle's RAC is a cluster of Intel X86 servers operating as a single processing resource pool with a single instance of Oracle's database. RAC offers load balancing and high availability and is a pay-for option with the Enterprise Edition (EE) of Oracle's DBMS, said to be four times the cost of Oracle's Standard Edition (SE)database.
RAC is included with SE but the cluster hardware is limited to 2 cores, which is a pretty noddy cluster. For any more cores, you have to pay the RAC tax on top of the more expensive Oracle EE software.
With VMware High Availability (HA) cluster, the server applications could run in virtual machines instead of physical ones and this also has load-balancing and failover, while being much less expensive because the SE database is a quarter of the cost of the RAC-using EE and you don't need the paid-for RAC option on top of EE.
Crucially, the virtual RAC is claimed to run faster than the real RAC, meaning VMware can say to Oracle users "Get more DBMS performance with VMware and stop paying the RAC tax and hiked EE cost."
The VMware virtual RAC is also said to have better management facilities than the real RAC.
So users with VMware HA can save a lot of money and Oracle loses that spend. VMware can't substitute for all RAC use cases, but it can substitute for a large proportion of them. It's said that RAC is fault-tolerant whereas VMware is high-availability. If you need full fault-tolerance, then RAC is the way to go. If you don't, then VMWare HA is a viable alternative.
Oracle has its own virtual server software (OVM), by the way, but it is said to be poor - less robust and mature - compared to VMware.
Oracle's general business strategy is to offer better-value Oracle database and application solution - hardware to application software stacks, from database to disk, as it were - by using commodity hardware and open source software underneath the Oracle software, which is most definitely not open source. It sucks cost out its stack to beat the competition, who look to make money on the hardware or software layers - or both, underneath the database and application products.
Crudely speaking, Oracle software is expensive, and what it runs on - the supporting hardware and software layers - is not. RAC is an exception in that clustered hardware and cluster software used to be expensive and RAC provided good value clustered server functionality for Oracle users as an integrated option for their Oracle software.
Now, VMware provides better performance for many RAC use-cases and is cheaper, exposing RAC's expense. What is Oracle to do?
One tack, it's claimed, is to oblige "customers to adopt Oracle VM Server or give up virtualization."
EMC high-profile bloggers, like Chuck Hollis and Chad Sakac, are telling Oracle customers using VMware to strongly pitch their views to Oracle about the need for it to fully support VMware, a kind of Oracle serfs' revolt.
Oracle has been gradually building HW-O/S-application stacks owned more and more by itself, such as the Exadata storage machine. The Sun acquisition can be viewed as Oracle's attempt to own even more of the stack by providing its own servers, its own storage, and its own O/S and system software in the shape of Solaris. The servers can be SPARC architecture, freeing Oracle from dependence upon Intel, and the O/S Solaris, freeing Oracle from dependence on any outside O/S supplier, including the open source community.
In this view, we're seeing Oracle build a walled garden for its customers, a locked-in database and application paradise, which will generate a huge recurring revenue stream for Oracle. If third-party software, such as VMware's, can knock a hole in the garden wall and let customers enjoy Oracle's software at much lower cost, then it's likely Oracle will react.
It looks as if the game-plan is for Oracle to sell an Oracle-controlled and built stack, from disk to database. Solaris has its own virtualisation technology, and thus, Oracle will have an in-house antidote to VMware. Whether it can steadily encourage its users to migrate off VMware and onto its own virtual server machines is an interesting question.
Very few businesses have taken on Oracle and forced Larry Ellison into a direction he doesn't want to go. He's a touchy guy, not a touchy-feely guy, and EMC's Joe Tucci will have a real contest on his hands if he tries to take Oracle on. ®