Slashdot Mirror
Oracle Is Latest To Take On VMware
BobB writes "Oracle is going after its piece of the hot virtualization market by introducing an open source Xen-based hypervisor to compete against those from Novell, Red Hat, and VMware. Oracle VM, unveiled Monday at the Oracle OpenWorld convention in San Francisco, enables virtualization on Oracle and non-Oracle software applications and on the Linux and Windows OSs. It also operates on industry-standard x86- and x86-64-based servers. Oracle claims it offers virtualization at a lower cost than competitors can." VMware stock dropped over 10% on the news; Oracle's stock rose. The market was not punishing Oracle for the unpatched zero-day vulnerability (public exploit available) that the company won't patch until Jan. 15.
I can't see the link between a Xen-based hypervisor and and a company being punished for a "unpatched zero-day vulnerability" that doesn't look like it is part of the hypervisor. Also, I can't see why the stock price would drop based on critical bugs. Stock prices should reflect number of people buying the software anyway. Hence Microsoft stock have value.
Please try to keep stupid statements like "The market was not punishing Oracle for the unpatched zero-day vulnerability (public exploit available) that the company won't patch until Jan. 15." out of the summaries. the market is NOT a technical forum, so unless this exploit can demonstrate some kind of loss for oracle, they have no reason to "punish"
If you mod me down, I will become more powerful than you can imagine....
It seems like everyone is trying to jump into too many new markets these days. Google with Android, OpenSocial, Microsoft with the Zune ... you get the idea.
Does have what it takes? I don't know, but it does seem odd that they have jumped into such a different market for them.
Seems a bit strange how RHEL 5.1 offers Windows virtualisation with Xen 3.1 and just days later Oracle does the same.
And how can this make VMWare stock drop by 10%? Xen ain't new (or great).
#include <sig.h>
Of course they cant. They can't compete with serious distro's and they wont be able to compete with serious Xen players. Thats not the point. wearing my tin foil hat, I'd say that their point is to fragment, or at least give the illusion to fragment, open source work. Oracle has lost a hell of a lot of real money to open source, and have been been brought to the enterprise open source table kicking and screaming. There is no money to be made here for them, they will gain little to no credibility in this space (not core business, blah blah) and they have never been known to have a warm spot in their hearts for open source in general.
People who think they know everything are a great annoyance to those of us who do.
for Oracle to get into the VM business.
Then they can ship pre-built VM images with oracle already installed and configured. Thus, the database server becomes a VM appliance (not quite a dishwasher yet...)
Easier to support (ie lower costs) especially if the VM runs Linux. As much as I hate Oracle, this following their 'legal theft' of RHEL it all starts to hang together.
However, it remains to be seen if they can build up their support side so that is basically 'sucks less' than it does now. There is a danger that they are spreading themselves too thin.
I don't think VMWare should get too worried by this. The overall market for VM's is huge. As long as the quality of their product stays high then their market will grow along with the overall market for VM Systems.
Sun is also rolling out a Xen-based virtualization solution called Sun xVM.
More info at http://opensolaris.org/os/community/xen/
This is a feature separate from Solaris Zones (OS virtualization) or
Brands (run Linux or Solaris 8 zones on Solaris 10) or hardware domains.
It's been my experience that the one thing Virtual Servers aren't good at it's io intensive applications like I don't know.... DATABASE SERVERS? At work we looked into virtualizing our development/testing environments the only thing we couldn't virtualize was the databases - too much of a performance hit. This seems interesting to me - why would oracle do this when they have fought the logical conclusion for so long - pre-packaged linux distro with their Oracle stuff built right in - deploy and go. Seems like one would be easier than the other.
Also - really can we get more retarded biased comments about stock prices in the summaries. It's good for a WTF chuckle.
Yep. That is why Red Hat made these t-shirts. Red Hat does not seem to have problems with Centos, but I'm not so sure about Oracle's copy of their distribution...
Yadda yadda, if Oracle doesn't fix it's own licensing policy than they still will be looking to take you hard for database licenses. They don't recognize software partitioning as a valid means of buying less licenses than there are CPU's in the physical box and when you run VMware in a cluster they want you to license your whole cluster.
http://virtualize.wordpress.com/
On this topic, Oracle also refuse to provide support for Oracle installations on VMware hosts.
I wondered why that was so, because VMware provides a common emulated hardware foundation, and logically ought to be *easier* to support than the wild variety of actualy physical hardware out there.
Anyway, now I know why.
Beware: In C++, your friends can see your privates!
I've managed to run Windows 98SE stably and reliably in Win4Lin 9.x on a Fedora Core host and VMware Server on a Debian host, though this is a desktop, not a server environment.
If virtualization can stabilize Windows 98, I'm sure that it can provide any help with stabilizing Oracle it can possibly use. If I had to run Oracle, I think I'd look for third-party support for Oracle and thumb my nose at Oracle Corp in the hope of getting more uptime than I can get with a native Oracle environment.
Any Oracle admins care to comment?
Tech Public Policy stuff
Oracle Is Latest To Take On VMware
Please - Xen does not a vmware copy make - vmware is so much more than a virtualization product, VMware are trying to make it THE datacenter management tool.
Alex
I'm too lazy to locate my slashdot login. If you post to me, I'll try to remember to check this thread tomorrow.
:
:
When Intel released VT technology, I experimented with a host OS-less VT implementation. It makes MUCH more sense than virtualizing on a *nix or Windows like Xen, VMWare, or all the others do.
A VM implementation is an OS within itself for the most part. It contains two major components and not much else. It contains
1) A process scheduler (yes, it's a fair comparison, the VT extensions make guest instances more or less a process with a different task switch mechanism)
This in itself can be accomplished in a VT aware microkernel, so the scheduler itself can be miniscule and more importantly can be much easier to manage and maintain vs. a full general purpose OS hosting a VM
2) Virtualized hardware access
a) Virtualized network occurs primarily at layer II, and half way into layer 3 for NAT, so a full TCP stack is irrelevant
b) Virtualized screens, this is easily accomplished by just providing a flat frame buffer (max 4megs) per guest OS
c) Virtualized USB is a bit trickier, this requires implementing the majority of the OHCI interface and typically requires that the stack is device aware, especially in the case of keyboards and mice since they need to be available to all systems. But the good part is, implementing a USB stack of this sort is actually easier than implementing a real USB stack. With the exception of keyboard and mouse, which could be implemented on a "different layer" of the USB stack, it would be mostly a matter of device ID mapping to separate machines. Shared devices would have to have a thicker emulation, but that's another problem.
d) Virtualized hard disk. This is much much easier than it sounds. The virtual machine would run its' own guest which would contain an iSCSI targer/router. Then the virtual machines would simply use a virtualized iSCSI host controller with boot functionality to access the block devices.
Well, I can go on an describe the entire implementation, but for the most part, I can safely say that I have either implemented as proof of concept or properly designed each function necessary to implement an entire VM. I used the source of a well known commerical microkernel for the basic design, but gutted everything from it that wasn't needed. A fully functional virtual machine manager with initial support for real time extensions fit into a 700K footprint and contained virtual screen, network, input support, and VERY limited hard drive support.
If I actually wanted to make a career out of it, I would search for funding, but system-level coding is my hobby and I don't want it to take over my life.
My learnings from the experiment were
a) Virtual machine managers are no longer rocket science. Intel and AMD have taken the rocket science out of it.
b) My implementation could easily fit in flash, and in known hardware configurations would easily replace the first level boot BIOS if a proper boot loader with made.
c) Much higher guest performance can be achieved by removing the fat of the host OS.
d) MUCH higher network and USB device performance can be easily achieved by replacing full protocol stacks with "layer II" stacks
e) Building a VMM on top of an OS is 100 or 1000 times more complex than coding it as an OS replacement. It should be an option on every machine and shouldn't cost more than $5-$10 a copy.
I also did another test which proved incredibly fruitful for pushing along UEFI adoption, I made a UEFI boot loader for the VMM and tested it on QEMU implementation with EFI. Since the QEMU version I had installed did not handle VT completely, it didn't work too well, but for the most part, it proved that by shipping a VMM as the "boot loader" of the system. The legacy BIOS included with the VMM wou
Looks like a nice opportunity to buy some VMW stock.
To be fair though, the market value jumped because they announced a new product. If those investors were aware of the total and complete 'shittiness' of their existing products (hell, 11 freaking versions and the interface still sucks donkey balls, unpatched exploits, etc...), they may not have been so interested in the new product.
Educating investors on what exactly they are investing in, and what the new product's likelihood of success is, is a very appropriate thing to do. Who wants to invest in a software company that is trying to muscle in on a preexisting field/market with a history of mediocre software design and poor maintenance habits?
-Rick
"Most people in the U.S. wouldn't know they live in a tyrannical state if it walked up and grabbed their junk." - MyFirs
All the major application vendors recognize that VMs are valuable and they also recognize that by refusing support for 3rd-part VMs *AND* offering a VM (even if it is someone else's with a new paint scheme), they can both make more money and squeeze competitors. Microsoft is already doing this with Exchange 2007 (note coming hypervisor/VM tech in 2008 server), and I expect other major application vendors might try the same thing.
This is clearly a monopolistic practice, but I don't expect our corporate overlords in the FTC to ever do anything about it.
One of the major selling points for VMWare is that it is ideal for development and if a major player in the development world, which Oracle is because they have the database, then there is much more appeal and legitimacy to doing virtualization through Xen. Now, personally I just got our company to buy a VMWare license for a couple of us because it is so easy and we're working with Macs. The issue wasn't a question of which VM system was better, it was a simple question of which was cheaper. I preferred VMWare by about a $10 margin over Parallels, and it turned out to be the less expensive of the two.
Now imagine thousands of people developing for Oracle asking themselves the same question, only now there is an official VM solution for Oracle development.. what do you think they will choose?
As to why it should reflect in stock prices, it's about name recognition. Most investors don't perceive Red Hat or SUSE or XenSource (Citrix) to be a competitor to VMWare, because they don't know who they are. Those same investors do recognize Oracle and as far as your average investor is concerned, virtualization is virtualization. Some of this may stabilize later as the difference in the products and the other competitors show up in stock market publishing, but don't expect it to disappear just because the people with the money realize there are other players on the field.
Back in my day when we chiseled our bits into stone and sent them by mule train from village to village...
You're thinking of VMWare, not Xen I suspect.
With a VMWare machine all communication with the processor must be interpreted, that is how the machine essentially works, it provides a fake communication channel that it can control. It does a really slick job of that, but it is still having to go through the extra step.
Not so with Xen, because it uses modified kernels that don't have to be interpreted (Which is why you can't just install your favorite OS in any old Xen like you can with VMWare.) This is why the overhead on Xen machines is 1% to 10% rather than 15% to 20% with VMWare. Essentially the only overhead you have is a minimal kernel that can not do much at all.
Further, VMWare has done some things recently that they say will put them in the same performance range as Xen if you're paying for their server product. I have not tried it.
It sounds to me like you've used VMWare and the defaults for a desktop or old server install and that means you used virtualized disks, which really, really sucks. If you're using VMWare player, you're doing the same thing essentially. IO most assuredly does not suck with Xen using drive partitions directly. Red Hat's implementation does something funky that way, but that may be why Oracle is making a big deal of theirs', it may be using the drive partitions directly, or even raw (without a format) as some databases, and I think Oracle does this, are prone to do. That means that they may be doing their own VM precisely because they can get better performance that way than with RH's product and if they can engineer the entire environment, I'd be shocked if their experts can't get better performance from a VM than most people setting it up even with significant experience.
Back in my day when we chiseled our bits into stone and sent them by mule train from village to village...
In 3. . . 2. . . 1
Why is Oracle getting into so many new markets? I think it adds value to some of their customers. The reason is service contracts. Many companies will not buy software unless they can also get service with it. What they don't like is having to buy 15 different service contracts and then watch the various vendors all say "The problem is in the other guy's stuff, ours works fine." These customers want a single point of contact for software problems and Ocracle wants to be that single point.
I used to work with older CDC mainframe equipment. It really was nice. There were full time CDC customer engineers always right there on site. I think Oracle and many others would like to be able to get back to providing this level of customer service.
There's probably more here than rebranding. Oracle has the resources to create their own Xen-based solution, and is certainly not going to rely on Red Hat's work.
That said, you're right to wonder at all the reaction to this announcement. Everybody and his dog are doing virtualization solutions, and the Oracle version is hardly groundbreaking. Indeed, since Xen only supports guest OSs that are hypervisor aware, it's not quite as robust as the "pure" virtualization that VMware does. And yet Oracle has managed to spin this as a solid challenge to VMware. Some marketeer really earned their pay on this.
Aside from the trendy corporate bashing that happens when any billion dollar crackpot-led company is on the menu, I lost all faith in Oracle the day they deprecated the term "database" in favor of "application platform", and added a couple zeros to the price.
Oracle, much like Microsoft, has lost its purpose. They absorb, invade and cheapen every single area of computing with their predatory acquisitions. It used to be that Microsoft was an OS company, and Oracle was a DB company. Oh, and Google was a search engine company. The lines have since been blurred and they're all trying to take over whatever's profitable.
Eventually we might see Oracle Burgers and Microsoft Shawarma. Anybody up for some IBM Pizza ?
-Billco, Fnarg.com
Our company uses VMWare simply because it's still the best virtualization platform I've been able to find for the x86 Linux platform and has the best upgrade path. We'd love to use open source but we've yet to see an open source product that can offer the same stability and features and host so many platforms.
When an open source competitor can run as reliably and fast on my dual quad-core Xeon server, can host at least Linux, FreeBSD, Windows 2000, Windows 2003 Server, and Windows XP, can do clustering with live migration, and do live backups then we'll seriously think about switching. Until then we're not interested because these are the features we need.
When an open source competitor can offer those though then there is a very good chance we'll switch. If VMWare wants to keep our business they either need to create new features we need, which is doubtful since this solution works great for us, or open source their code. We'd pay for support ven if their code was free.
At what price learning? At what cost wisdom? The price is a man's peace of mind, and the cost is his life.
middle wear is a belt. middle ware is software.
Oracle tried to come out with a linux distribution too. Do you know anyone who uses it ? It was so lame. They even forgot to remove all the "Red Hat" screens in the install. They have been darn successful with databases. They should stick with that and stop trying to put fingers in pies that they know nothing about.