Reasons why not to use Oracle VM…
I get a little bit cranky about some statements posted on the (Oracle maintained?) blog "Oracle VM Blog". Not because they are not true, but more about people tend to use them out of context. A lot of people I know state afterwards, after reading those pointers, OK, so "Oracle VM is 3 times faster than other virtualization products". They do not read what it actually says (also now on the Oracle VM Blog). It states: "Greater efficiencies—Three times more efficient than the other, leading server virtualization product." (aka VMware?) or "Three times greater efficiency than current x86 based server virtualization products".
Here we go (again). It depends...
First off all Oracle VM, as far as I currently know, is based on the XEN hypervisor. As far as I know this is / was an open source project, but lately Citrix Systems acquired XenSource (22/10/2007), and part of the project split and moved to the Xen project . Essentially Oracle VM has the same characteristics as the Xen hypervisor. Said this, this means that it has (at current state) the benefits and the lacks of the Xen hypervisor (v 3.1.1). Xen was initially developed with only the possibility of virtualizing Linux environments in mind. MS Windows guests are not so great regarding performance in comparison to "other" vitualization products.
In the Oracle FAQ whitepaper on www.oracle.com it clearly states:
Will the performance of Microsoft Windows running on Oracle VM be similar to the performance of Enterprise Linux running on Oracle VM?
No. Currently, Microsoft Windows runs significantly slower on Oracle VM than it does on native hardware. However, Oracle is developing paravirtualized Windows drivers that will substantially improve the performance of Windows on Oracle VM.
So, if Windows guests are important to you, you might want to wait.
...but said that, something has radically changed since Oracle VM. Oracle know fully undoubtedly supports a virtualization method and this is only good news, especially because I really love the possibilities given by virtualization technology, that I as a private person, in terms of hardware, couldn't effort. Nowadays I can actually build a RAC system with low cost hardware.
Oracle VM comes in two parts: Oracle VM Server and Oracle VM Manager. The following statement in the Oracle VM FAQ whitepaper,
"Consisting of open source server software and an integrated Web browser-based management console"
gives me the reasons to believe that Oracle will contribute to the open source Xen community via co-development of the Oracle VM Server part, but that this could not be the case regarding Oracle VM Manager. In my mind, this would be as expected because I would integrate the manager part in Oracle's flagship Oracle Enterprise Manager Grid Control. The disadvantage of this would / could be that you would be licensed bound (and OEM grid doesn't come cheap).
So, all in all, when put my mindset in perspective, using Oracle VM would be great when your environment is dedicated to Oracle solutions and fully linux based. You get the advantage of a fully by Oracle supported environment, cheaper and more powerful than other virtualization (non-Xen) products. Because Oracle has a great development team behind the Oracle VM product, it probably will out run Xen competitors in the end as well (but currently I can't find hard proof on the internet for this). One thing is currently clear, the Oracle VM virtualization support is cheaper than the "other" virtualization market competitor.
I breath, eat, swim, love, am nerd-ish about Oracle software. I support people already since 10 - 15 years as an Oracle DBA (and still love my job), but I think that people should realize that every great solution comes with a prize. If you try to make a consistent, well thought through, maintainable and balanced decision, taken into account the pro's and cons, then you are more or less sure that the solution will fit 99% of your needs. Although you will never gain up to the desired 100%. This is the daily part of dealing with software (issues). You can always advice the customer to try a "ctrl-alt-del" approach and see if the problem has solved. In the end it is stuff made by us humans, so error prone.
Marco
Probably overkill, but this is my personal website. Views given here are my own and maybe do not reflect the general opinion of my current employer.

January 11th, 2008 - 23:01
glad to stumble across your blog, Marco – yet another nice Oracle-related blog !
BTW, I’m the guy (matelot) who posted on XMLDB@OTN the question you put on the page : http://www.liberidu.com/blog/?page_id=241
The XMLDB Repository
How is DefaultTable populated after dbms_xdb.createResource? (2007)
January 12th, 2008 - 00:10
Yeah that was a great OTN post where Mark explained in more detail the inner workings of how stuff works together.
I hope you will find here a little info for along the way.
My English was really crappy though on that forum post (the only excuse that it was almost 01.00 am)
September 19th, 2009 - 06:03
half of what you state is wrong, which seems odd when talking about unclear claims by oracle – though i agree on that.
“Xen was initially developed with only the possibility of virtualizing Linux environments in mind. MS Windows guests are not so great regarding performance in comparison to “other” vitualization products.”
- Windows was among the first paravirtualized OS, it is just by M$ decision that it never got made publicly available.
- Xen was designed not primarly with Linux in mind, but with extremely large, world wide grids in mind. Small difference, if I may say.
I’ve been running VMWare since the 1.x versions and Xen since it’s 2.0. trail, and after some years with Xen I recenctly put a lot of time in ESXi and the likes.
If I were to virtualize a huge windows shop i’d probably go with VMWare too, but performance-wise it’s a total nightmare if you are used to Xen (in it’s _slow_, HVM mode)
Still, a “huge” windows shop of 500-1000 VM’s is far from what Xen was designed for. So if you’re looking at deployments of Amazon’s size, you’re obviously looking at something like Linux, which is where you’re statement looks right. But for different reasons that you seemed to assume.
About the Citrix bit:
None of the big Xen shops give a damn about Citrix and their marketing buzz. Citrix, as of today is still assigning random mac addresses not assigned to them to the VMs in their breed of Xen.
Anyone running a *lot* of VMs is not using Hyper-V, not even VMWare and most certain not Citrix XenServer.
From what I’ve seen, most tend to do homegrown stuff because none of the commercial offerings are really “enterprise grade”, and the vendors seem extremely clueless as how to get there.
in my opinion, the reason why VMWare will be leading the virtualization market for much longer is simply because of their experience and the real production usability of ESX. There’s not even another product with a real NPIV implementation on the market right now, and none of them have even come up with a SNMP mib…
Last, as you said Oracle VM will be really good for Oracle&Linux shops.
You might want to think about the following question:
Why should a company use anything else than this combination for their smaller oracle deployments?
September 23rd, 2009 - 13:42
Although anonymous, thanks for the additions in supporting my ideas, at least that what I make of it,and putting some “dots” on the “i” I wasn’t aware of. Don’t be to harsh, this blog post was once created while the “pro marketing Oracle VM campaign” was at its highest and I was really cranky writing this post.
The problem with VMware, from an Oracle DBA viewpoint, is that Oracle doesn’t maintain a simple and clear license strategy for those environments, but a very clear one when it is situated on Oracle VM.
It depends what you call “small”, we Oracle guys, most of the time don’t have any “small” environments anymore…
Also, in the meantime the whole market has changed; Regarding virtualization products Oracle (after the Sun deal closes) has Oracle VM, Virtual Box and the virtualization products of Sun…