Diggin' technology every day

November 14, 2011

Oracle throws in Xen virtualization towel?

Filed under: Virtualization — Tags: , — Nate @ 7:03 am

This just hit me a few seconds ago and it gave me something else to write about so here goes.

Oracle recently released Solaris 11, the first major rev to Solaris in many many years. I remember using Solaris 10 back in 2005, wow it’s been a while!

They’re calling it the first cloud OS. I can’t say I really agree with that, vSphere, and even ESX before that has been more cloudy than Solaris for many years now, and remains today.

While their Xen-based Oracle VM is still included in Solaris 11, the focus clearly seems to be Solaris Zones, which, as far as I know is a more advanced version of User mode linux (which seems to be abandoned now?).

Zones, and UML are nothing new, Zones having been first released more than six years ago. It’s certainly a different approach to a full hypervisor approach so has less overhead, but overall I believe is an outdated approach to utility computing (using the term cloud computing makes me feel sick).

Oracle Solaris Zones virtualization scales up to hundreds of zones per physical node at a 15x lower overhead than VMware and without artificial limits on memory, network, CPU and storage resources.

It’s an interesting strategy, and a fairly unique one in today’s world, so it should give Oracle some differentiation.  I have been following the Xen bandwagon off and on for many years and never felt it a compelling platform, without a re-write. Red Hat, SuSE and several other open source folks have basically abandoned Xen at this point and now it seems Oracle is shifting focus away from Xen as well.

I don’t see many new organizations gravitating towards Solaris zones that aren’t Solaris users already (or at least have Solaris expertise in house), if they haven’t switched by now…

New, integrated network virtualization allows customers to create high-performance, low-cost data center topologies within a single OS instance for ultimate flexibility, bandwidth control and observability.

The terms ultimate flexibility and single OS instance seem to be in conflict here.

The efficiency of modern hypervisors is to the point now where the overhead doesn’t matter in probably 98% of cases. The other 2% can be handled by running jobs on physical hardware. I still don’t believe I would run a hypervisor on workloads that are truely hardware bound, ones that really exploit the performance of the underlying hardware. Those are few and far between outside of specialist niches these days though, I had one about a year and a half ago, but haven’t come across one since.



  1. The bigger question will be do they still treat their version of virtualizations use of soft partitioning as “hard” versus all the other hypervisors when it comes to licesning?

    I found it utterly shameful that Oracle would treat their Xen model which used soft partitioning for CPUs as a hard partitioning scheme for licesning, but didnt for VMWare, Hyper-V, KVM, or even XenServer (which is still Xen). The big issue here is the cost of licensing for database is drastically different when you don’t have to license all the cores on the host, and only the vCPU’s given to the DB server.

    Comment by gchapman — November 14, 2011 @ 9:10 am

  2. Yeah that is true, the hardware based licensing is a double edged sword. If your rich enough to afford Oracle EE then licensing costs probably aren’t a big factor since your going to use it regardless. Oracle SE on the other hand with it’s per-socket licensing is quite nice. You still have to license for all of the CPUs on the box which is good and bad. The bad is obvious if you only need 1 vCPU you may need to pay for 2-4 sockets (x number of boxes if you plan to vmotion a bunch). But the good is you can have unlimited VMs on the host running on the same license.

    Last time I used Oracle was several years ago but licensing a single socket of Oracle SE + single socket of ESX 3.5 (back when VMware officially did not support single socket systems) on quad core servers (dual socket, but only one in use) saved my company quite a bit of licensing costs, allowing us to run multiple separate Oracle instances at the same time in a more isolated fashion. We went from what would of been at least 5 Oracle licenses down to two.

    I feel your pain if your on EE though…the licensing for that is truly harsh.

    The bigger issue with Oracle and VM though is the support. I think they still don’t officially support the likes of Vmware running under Oracle, forcing customers to either beg, or reproduce the problem on bare metal (or Oracle hypervisor) before they will assist which can be a major sticking point. I never had a problem running Oracle in Vmware going back to 2007, though our production Oracle system at the time was on bare metal, partly for support, more for performance – Vmware didn’t really give us any advantages over bare metal for that system at the time, and CPU overhead at the time was of course higher than it is now.

    I’m about to deploy another (very small) Oracle Standard Edition instance to support vCenter on my new vmware project. Since it is vCenter I opted for named user licensing, and even went for yearly license instead of perpetual, really got the cost down, and lets me run my vCenter database on Linux instead of windows 🙂 (I don’t have any background in DB2 so wasn’t comfortable going with that).

    My company’s main product that they use to support their business is apparently license per VM (will try to confirm it again), which to me would suck. I’d much rather pay a higher cost and license on the hardware(depending on how much higher cost of course). Being able to have unlimited instances would be very handy, especially for an application that doesn’t scale gracefully.

    Comment by Nate — November 14, 2011 @ 9:40 am

  3. While Oracle doesn’t “officially” support VMware, they will not refuse to give support on Oracle systems running in a virtual environment. I did attend a talk by the head Oracle guy from VMware and they have their entire Oracle instances virtualized and had mentioned that there has not been a single issue that Oracle had a customer recreate on physical hardware in order to garner support, this he said went back to 2007.

    There are some good whitepapers for Oracle/VMware integration etc:

    that said, its still cost prohibitive for EE users (like my group) to run on our ESXi boxes compared to running on our IBM Power servers running AIX.

    Comment by gchapman — November 14, 2011 @ 10:44 am

  4. I hear ya, I remember going out of my way to buy dual core processors even after quad core had come out because dual core had better bang for the Oracle EE licensing buck (much higher clock speeds).

    Comment by Nate — November 14, 2011 @ 10:57 am

  5. Hardware virtualization support a la VT-D and hypervisors may have shrunk the virtualization overhead, but having one separate OS instance per VM is inherently less memory- and cache-efficient that something like Solaris 10 Zones. Zones may not be as versatile as full-blooded replication, specially in heterogeneous and legacy environments, but they are a superior alternative for Solaris shops.

    Oracle has far too many virtualization offerings, between Zones, hardware containers on the SPARC platform, Xen, VirtualBox, and I suppose you could count KVM support in the OpenIndiana fork of Solaris.

    Comment by Fazal Majid — November 14, 2011 @ 2:48 pm

  6. thanks for the post!

    I had no idea that Oracle had forked KVM as well. I totally acknowledge that the concept of zones has less overhead than a full blown hypervisor, it’s all about trade offs. The overhead in many cases doesn’t matter though. Which is why the likes of user mode linux, freebsd jails and even zones (reminds me of that reseller named ‘Zones’!) to some extent haven’t really shown to be any competition to the hypervisor approach (at least from a market share perspective).

    I am not surprised that Zones is more optimal for Solaris shops which is why I think it may be difficult for Oracle to convince non Solaris shops to jump ship to Solaris from Linux or other platforms just for Zones when there are things like vmware, kvm, hyper-v.

    I suspect part of the issue Solaris faces is outside of Vmware, and Oracle’s own VM offerings are there any other hypervisors out there that run Solaris? I don’t see (official at least) support from Red Hat or SuSE KVM, Citrix Xen, or Hyper-V.

    Comment by Nate — November 14, 2011 @ 3:06 pm

  7. Honestly, it’s not the cost of licensing from Oracle that blows me away, its the support contracts. For what we pay annually in support, you could pick up a pretty nice 3PAR rig.

    Comment by gchapman — November 14, 2011 @ 3:51 pm

  8. Can you disclose what you pay from a “% of original license cost” perspective? I don’t remember what my company paid, I went through two different Oracle license audits at one company, the first one I was still a newbie to the company and didn’t have much input, the 2nd one I was much more involved and that’s when they agreed to my plan to move to Oracle SE. I still have all of the emails from back then, I would be curious to compare, assuming I have the quotes in an email attachment what were paying vs you. Not that we had strong purchasing power by any stretch. But I’m curious how much it might of changed over the years.

    What specifically is it about Oracle EE that keeps you on Oracle EE ? Do you really need to use features like partitioning and stuff? For my company at least at the time it required no code modifications to our applications at all to switch to SE. It required changes in our backup procedures because you could not(at least at the time, maybe it’s changed) run RMAN from a standby DB on Oracle SE, but I came up with a workaround for that too. Also we had to come up with a workaround to the loss of transportable table spaces which we used for reporting snapshots, fortunately the workaround for that wasn’t complex either in our case.

    Comment by Nate — November 14, 2011 @ 4:03 pm

  9. Oh and I have long thought that Oracle users should, if at all possible use Oracle SE and use the cost savings from that investment to invest in some really good storage 🙂

    My company, at the time once we switched to Oracle SE, had 3 different physical servers, each with a single quad core CPU in it (this was 2007), as you can imagine the licensing costs of 3 x Oracle SE licenses vs the equivalent licenses for 12 CPU cores of Oracle EE. I think before we migrated to Oracle SE we had at least four database servers(probably 5), each with at least two cores. Part of the migration to Oracle SE involved new hardware to leverage quad cores, it more than paid for itself!

    Our DB consultants, at the time had a ‘standard practice’ of installing Oracle EE + partitioning (used by their monitoring) they did not realize initially the impact that had on our license audit. So another thing they had to do when we went to SE was they had to re-work their monitoring so it did not use partitioning, they were going to do it anyways since other customers were making the migration to SE too. My company had originally licensed Oracle SE “One” (the really low end one), but apparently didn’t communicate well with the DB consultants who went right ahead and installed Oracle EE. Oracle support took notice fairly quickly after we started filing support requests..

    Comment by Nate — November 14, 2011 @ 4:08 pm

  10. I’ve learned a lot about what not to do based on how we implemented Oracle. We are on year 4 of a 1 year rollout to EBiz. Our IT “Director” at the time thought he was getting a sweet deal when he got the entire EBiz suite at a pretty steep discount; of course, the money is always made on the support contract. I believe its running at about 22% annually, suffice to say IT Director is no longer an employee.

    We are deploying GRC, OBI, EBS, Hyperion, Vertex, and a few other components in our deployment. This runs on EE for the database tier on our Prod P550 boxes running AIX 6.1 which is licensed by the core. On Development we went with named user which was easier (and way cheaper). So for what we wanted to implement, SE was not an option. I think in hindsight there would be a lot of things we would have done differently.

    Each Oracle implementation is its own can of worms. Some companies just need the DB, others need everything under the sun. We are replacing our HR, Accounting, Invoicing, Order Entry, CRM, etc. It’s the entire business running off of one unified system, and as such, there have been major pain points that were not foreseeable (well they were but that’s a whole different story). The really sad thing is, that all of the hardware I had to evaluate, deploy and configure for this entire project hits its initial 3 year support contract right before the system is supposed to go live. A lot of it is under powered in my view for the newer versions of some of the apps. JVM requirements alone for most of the GRC line is 16GB of RAM per box, and that doesn’t include overhead for any of the other components. The really scary thing is, there has not been a valid performance benchmarking plan put forth, so there is no way right now to determine if what we have in place will actually be able to support the number of users we need it to.

    If this was my project, that is what would keep me up at night.

    Comment by gchapman — November 15, 2011 @ 3:11 pm

  11. ouch! That is too bad, crazy. What a mess!

    Comment by Nate — November 15, 2011 @ 3:20 pm

RSS feed for comments on this post.

Sorry, the comment form is closed at this time.

Powered by WordPress