It is perhaps not obvious why VMware, the paragon of proprietary software profits, has become such an open campaigner for open source, open standards, and open APIs. What has changed to make openness a sound business decision for a company that continues to mint money with its various closed-source products? I've spent a fair …
The cloud vision
Let me get this right. VMware's vision of the future is to have a database, an app server, a message queue and so forth, all running on top of a "hypervisor" that mediates shared disk, memory, cpu resources?
Can anybody remember what the definition of "multi-tasking operating system" was?
D'oh, because we're reinventing the wheel.
I realise your question is rhetorical but ...
this is not to do with terminology or technology. If VMWare can make out they've invented all this stuff and enough people believe them then they clean up.
Look! New and shiny! Roll up, roll up.
Less cynically, VMWare will tell you that all their stuff is better than the old hat (red or otherwise) you're running to do the same thing and they may be right but marketing BS is all it really is.
Traditionally you could run a large number of services on a single host OS... However, through a combination of security concerns, poorly designed operating systems (mostly windows - lack of chroot) and poorly designed applications, it's no longer fashionable to install multiple services on a single OS.
Virtualization (yes its been around a long time - often as a glorified chroot), was primarily useful for hosting environments where each customer could have root on their own virtual environment.
I personally am not a fan of vmware for a number of reasons...
Proprietary management/console tools, tying you in to windows (they used to support linux but not anymore)... By contrast, KVM can be managed via SSH and VNC. Also seems rather foolish to create a dependency on windows, when microsoft are trying to kill them with hyper-v.
Expensive - they give you the base hypervisor for free, but its pretty useless on its own... any remotely useful features are nickle&dimed... Considering most of the competitors are free, vmware is massively overpriced.
Flakey - vmware esx seems to use some pretty crufty hacks to improve performance, but these are only tested on specific guest os... If you want to run something else, or build your own custom kernel etc you can have all kinds of problems.
Questionable licensing terms - their EULA prohibits benchmarking, why would you do that unless you have something to hide? The obvious reason is because vmware performs very poorly compared to its competitors, and they don't want that fact made public.
Their stated reason that they don't want erroneous benchmarks put out is ridiculous, their own benchmarks are highly likely to be erroneous too - in order to make their product look better... What users need is a range of benchmarks from different sides so users can draw their own conclusions.
Personally i feel vmware have no future, they are the new netscape... Sure, they were a pioneer in virtualization on x86 and were once the only game in town and could rake it in. Now, they're a dinosaur, charging for a product everyone else gives away for free...
They will lose windows customers to microsoft, as hyper-v comes with windows, and will offer better support for windows (and who knows, ms may even modify windows so as to cause problems when running in vmware - wouldnt be the first time)...
They will lose linux, bsd and solaris customers to kvm and xen, they have already alienated the linux users by dropping support for the console and forcing users to run windows management hosts...
They won't make much of a splash in the cloud hosting arena, because the idea of the cloud is to scale and that doesn't work when your paying through the nose for software...
The problem is businesses depend on a *huge* amount of legacy Windows software, badly written by self-taught bedroom "developers" using pirate copies of Visual Studio, which can't handle not running as Administrator and for which the Source Code is lost.
The only way to make that heap anything like secure is to run each process in a separate VM, so no process can interfere with any other process. Ugly? For sure; but it'll drive down the price of the hardware needed to make it work. And that hardware will *fly* with a proper multi-tasking OS!
kosh, that's exactly right, because hypervisor != OS. The modern use case for virtualization is to move disparate workloads adaptively around one's hardware infrastructure, as well as compartmentalizing applications. You could do the same with careful OS configuration and BSD jails/LVS, but you'd lose a lot of flexibility and the ability to move workloads seamlessly.
If there wasn't a compelling business case to do this, VMWare would be worth a lot less. It's also why these features are never included in their open source/free offerings.
KVM is more competitive than VMWare
Matt's comment about RedHat needing KVM to be more competitive in this article caught my eye. I've been running a private cloud built on vSphere for several years, and earlier this year took the decision to phase it out and move to KVM, for a number of reasons.
KVM is able to leverage all of the underlying capabilities and qualities of the Linux kernel, and we found it to have matured to the point where VMWare comes out second best in terms of robustness and performance and hardware support. So far, KVM is proving to be lower-cost to operate too, especially as we've encountered serious costs with VMWare due to compatibility problems with leading server manufacturers.
Factor in VMWare's ill-judged price increases this year, and KVM is more competitive for those organisations who have the in-house skills to migrate to it.
Yeah who really cares.
I sure don't care what VMware is up to anymore. They were impressive years ago, but these days they are irrelevant. KVM is much better and more flexible to use.
I can file bug reports on KVM and be listened to. VMware won't even accept a bug report without a support contract. And even if you manage to give them one, they don't really seem to care. Who needs them.
And yeah who needs another OS, which really does seem to be what they are trying to turn the hypervisor into. What's wrong with the Linux OS sitting underneath their hypervisor in many cases?