I don't think their hold is nearly as effective as Microsoft's once was. The hypervisor API is very simple compared to the OS and lock-in is practically nonexistent. It's also not nearly as common - how many machines are running hypervisors these days?
Actually, VMware is more like Palm - they started something disruptive and defined the market, but somehow lost the edge (Cloud Foundry is two years after Heroku, four years after Amazon, etc).
Exactly. VMWare, right now, has the best tools overall. Everyone else is playing catch-up.
However, everyone has caught up enough that there are plenty of choices for competent hypervisors. VirtuaBox, Xen, KVM, Hyper-V - they may not be enterprisey enough yet, but they work fine for more and more cases.
I was chatting to my friend who worked for MS NZ in their sales team and we were talking about the usual "who's bigger, who are the main competitors" type stuff. He asked who we all thought MS' biggest NZ competitors were. Naturally I answered Oracle/Sun as one of the main ones. He just laughed though, then told me that it was VMWare. If he knew, then MS are surely pretty keenly aware of this.
Microsoft was about taking over people's hardware (i.e., protecting Windows' installed base), VMware is about helping people utilize their hardware to its full potential. It turns out that one of the fundamental thing we want to do with modern workstation and server hardware is run multiple OS instances on it simultaneously.
This surely sounded subversive, scary, illegal, and weird to MS leadership who expected users to do web, games, mail, and office apps on their hardware. MS required a per-instance license anyway, so how could they lose?
That said, I've heard good things about MS's virtualization technology, aside from the little problem of it requiring Windows for the base OS.
Well it is more subtle than that. No sane organization runs multiple OSs for the sake of it. The ideal number is two or three: it hedges your bets against one OS vendor doing something unfortunate and yet keeps your support costs down. So why do you want to run multiple OSs on the same hardware? Because your apps already exist on those OSs. It's all about the apps, always was, OSs are just plumbing. And you are completely wrong about MS: they fully expect people to be running SQL Server, IIS, Exchange and the rest of their back-end on Windows too. The problem always was, it's difficult to run all those together on one Windows instance which is why you would have lots of boxes. A hypervisor running n Windows instances, each of which running one MS server product is an ideal situation for MS.
Any nontrivial organization around for a sufficient period of time will end up with a collection of old servers running various apps. The accounting system. The document management solution purchased 5 years ago. The sprinker system. The program that's needed to recover data from the old backup tapes that everyone hopes will never be needed. And so on.
Many of these apps can never be migrated to newer OSes, nor can they be switched off (yet). These are often good candidates for virtualization.
A hypervisor running n Windows instances, each of which running one MS server product is an ideal situation for MS.
OK. But in the real world we deal with systems other than the ideal situation for MS.
That's not really their marketshare/growth area. The real money is the enterprise stuff, and no one is as slick in there. Having hot swappable VMs running on blades connected to a SAN with real time load balancing is as slick as hell, as well as costing a pretty penny.
Add to that a really slick management console, and there's a reason why they own the space.
I would question their long term (15-20 years) viability. The major part of their enterprise profit is for legacy apps, virtualizing some fox pro database backended windows 2000 box. It's a lot easier to virtualize the thing than reengineer the thing. Sooner or later though, the requirements for all systems go away, and stuff like that will get migrated to some newer system, cloud or otherwise.
The market for straight cloud systems, like s3, is a lot more even.
I really don't think it's just for legacy apps - people install VMWare as the base platform for enterprise data centers - it was a pretty smart move for EMC to acquire them as it makes it a bit of a no-brainer to then have EMC SANs as well which is most of the money typically goes.
VirtualBox is a great tool for sure. VMWare does beat VB for large amounts of virtual machines. But that should really not matter to anyone who's not running a datacenter.
In the personal use realm, VB (mostly the guest tools) just isn't very slick compared to VMWare. Dual screens usage, seamless/unity modes, resizing windows, clipboard transfers, drive mounting, etc, are almost always a better experience with VMWare (fusion/workstation). The biggest reason I use VB in many places is because you can't argue with the price.
Creator of Vagrant here. I'm sorry you had a poor experience.
Get in touch with me at mitchell.hashimoto@gmail.com and I can help you get through any issues you may be having. Not being able to set it up easily is definitely a "bug" in my book!
Not sure I would agree complete with this assessment. Reason Microsoft was able to get that kind of stronghold was due to applications(Word, Excel - though they copied these - and others vendor apps/drivers written exclusively for windows). It is pretty much the same reason why desktop Linux is not able to make much inroads into the windows desktop OS market(it's the apps). Of course Apple and Google are trying to just bypass this market altogther(handheld, web). But it is the apps those matter and not sure where does VMWare has the apps to lock folks into the VMWare?
OS Virtualization is necessary because the software stack has got very complicated. This is true not only of Windows, but on Linux as well. A working server has moving parts from web config files to database configuration to PAM. All these make redeployment from scratch a painful experience. It is often cheaper to just copy a working deployment and tweak it.
Microsoft, in particular, has made this worse by keeping configuration details in registries and not clearly identifying which keys should be exported for an application to continue working.
Minor gripe, just because the contents are stored in the registry doesn't preclude them from being exported or clearly identified. The clearly identifying anything I entirely agree that Microsoft has done a poor job of.
You are right. The purpose of registries is for components to register and announce their location. If a complicated piece of software has a lot of dependencies then no amount of registry-fu will solve the problem of cleanly migrating the application to another machine. Virtualization is the only sane solution.
