When is an operating system not an operating system? It used to be the case that the role of the operating system (OS) was pretty well defined as a layer of software responsible for controlling the use of and access to physical machine assets such as CPU, memory, disk, network, and so on. As the industry has evolved, however, so too has the role of the OS. Today, for example, …
At least, it probably will be in the desktop world.
Think about how easy it would be for Google to own your computer as well as your data this way (I think this is the idea behind Chrome OS).
I am quite excited about this prospect, not about Google, but the fact that I may not need to go through the process of installing an OS, but instead just install the hypervisor. I am intrigued to see how companies like MS will use it to fight piracy etc. I am also interested to see how the open source community would work with the concept - after all, open source is based on the ideal that its just software. Could the open source community support this?
Surely, an Operating System is a software layer that abstracts the hardware so that application programmers do not have to worry about physical I/O and can use logical I/O. There are two areas of Virtualisation : servers and the desktop.
Virtualisation software like Xen and VMware have been created to provide Virtual Machines with virtual I/O, then it is logical to run an Operating System to manage the interaction with the Virtual Machine otherwise the application has to worry about interacting with the virtual hardware.
However, if Xen included the option of providing a JVM instead of an x86 virtual machine then application developers could write in Java and run them without the necessity of a OS to host the JVM as it is provided by the Xen in this case which is acting like an OS.
If Xen was also extended so that WINE was incorporated so that it could create a WINE VM, Windows applications could be run without the need to have an instance of Windows. Similarly other OSes could be made redundant with similar technology.
However on the desktop, Windows already provides a degree of virtualisation using WOW (Windows on Windows) to provide backward compatiblity. Other OSes can use WINE to achieve similar results. Similarly VirtualBox provides the virtualisation functions to allow muliple OSes to run on Windows, Linux, OpenSolaris and Mac OSX using Virtual Machines.
Until application programmers can target the Virtual Machine environment through a standard API then the OS is here to stay. However, if a JVM was natively supported by Xen and VMware then it could happen much sooner.
OK, so you want a lean OS for efficiency in operation in a VM, and you want a standardised, non-customised OS for efficiency of support.
To me, that sounds like what you want is a microkernel OS. Microkernels are by nature lean and are dynamically self-optimising, giving you the best of both worlds.
The world and his dog has been crying out for a decent microkernel OS, but no-one has ever been willing to put the time in, instead developing more and more convoluted ways to patch, modify and blu-tack external code onto the major monolithic kernels.
If virtualisation finally provides the impetus for a truly world-class microkernel architecture, then virtualisation is indeed the way of the future.
Even on the simplest systems (small embedded processors), the "OS" is typically divided into various bits providing various services. At the bottom you have the basic kernel, which gets calls from apps (or whatever you choose to call discrete software modules with defined interfaces) to send a message and then makes the appropriate calls to other apps to send them the message. Then you have timers, interrupt handlers and low-level drivers for ADC/DAC/logic-level/other peripherals, all of which need to do register-level operations to control real hardware. Then you may have high-level drivers to convert from real-world values to ADC/DAC counts. And finally you're into application-level processing.
So which part of that is the OS? The answer really is "everything that doesn't do application-level processing". It's a pretty vague question, and appropriately enough gets a vague answer.
But which part is the OS *kernel*? That's a very specific question, with a specific answer: the section of code which directly controls activation and deactivation of all other sections of code.
Good point - which raises the question, what is virtualisation *for*?
As far as I can tell, there's three theoretical reasons for running VMs:
1) You want two different platforms (eg Windows and Mac) on the same hardware.
2) Sandboxing of functionality, so if one fails or is compromised the others are unaffected.
3) Efficient use of hardware, so that power usage is minimised (when less processing is required, some hardware can be 'put to sleep')
In fact, only option 1 is where VMs are truly needed, and they're required to host the guest OSs.
However, I suspect that the real reason for much of virtualisation seems to actually be as workarounds for legacy design limits, such as numbers of TCP/UDP ports and poor OS multitasking performance.
There are plenty more reasons to have visualization, among them being space (physical), fail-over, redundancy of core servers, the ability to easily and quickly back up a whole computer, and then restore it independent of hardware.
On top of all this there is the management options with the ability to remove problem servers from networks, commission new servers relatively quickly from spare capacity to suit needs.
Of course there are issues with it, like memory swapping etc, but overall for a lot of my companies clients (we are recommending to more than 95% of the at least partial visualization of their server environments) visualization is a viable option for reducing costs, centralizing and consolidation IT infrastructure and providing a future proofed solution.
Testing is a big one, actually is a huge thing in all of IT.
And is often not really given the correct due, because of course you can create something without or with minimal testing.
The HyperVisors came into their own really for browser testing, across operating systems, and on same operating systems but different browser versions. Though back then when personally confronted with that problem, I still opted for making everyone run a different browser, but on reflection that was probably not the right move, live and learn I say.
Testing is now crucial, or at least should be championed. Starting to see more people take on this idea, but it is still something that needs a hell of a lot more work done on. Testing crosses not only quality control, but also security, and it feedbacks very obviously into development, and it still doesn't really have the respect it deserves in business or development, and it is up to development to make the business case here.
Another good place for HyperVisors is honeypots, but not honeypots for analysis of attacks, honeypots for misdirection. A lot of systems will have spare CPU cycles, throw a visor on and you can start to pretend that system is three systems, now if someone tries to compromise the chance of hitting the real system is now 1 in 3, and that is a measurable bit of security. If a system that is fake is compromised it is easy to detect that has happened, so it gives a heads up as well, useful IDS.
And yet another useful use for them, here in a distributed fashion, is to activate the distributed hypervisor in down time across a company, where they may wish to do some major distributed number crunching.
Day to day distributed use is still quite dodgy, you can admin many systems quite easily from one point anyhow.
Overtime as the visors start to prove themselves then perhaps people will distribute visors across multiple systems, perhaps ARM based, quickly turn on another system, and attach to the visor, and programs could start taking advantage of that.
But not all systems in an organisation under one visor, those who do that will probably live to regret it, there is just too much that can then go wrong, too many eggs in one basket and all that jingo.
By definition, the OS controls the hardware. That's what "operating system" means. Call it an OS, or a kernel, or the monitor, or a hypervisor, or what have you. The logical unit involved controls the hardware. This is part of the reason MS/PC-DOS wasn't really an OS, it didn't control the hardware ... DOS, and all versions of Windows prior to NT and it's derivatives were program loaders, at best.
Anything that looks like an OS, but isn't in total control of the hardware is emulation, a virtual machine, a program loader, or simply an executable or collection of executables, hopefully operating in harmony (for example, the user interfaces in Mac OSX, KDE/Gnome/XFCE on Linux, Win2K thru' Win7, etc). To suggest otherwise is silly.
Likewise, the ton of shovelware that comes with Ubuntu et ali doesn't make it a "different" OS. Rather, the fact is that the actual OS under Ubuntu is the very same Linux OS that comes with Slackware or Puppy (although configurations can change, the core OS is still the same).
I have to join the "OS drives the iron" camp. A hypervisor that sits directly on the iron is the OS even if all it does is serve up virtual images of the iron, VMs. The VMs are then just standardized APIs, if you will, that run programs even when those programs are what is typically considered an OS. I'll let others argue about what functionality "needs" to be built in what.
What I want to know is what is the OS when a hyperviser is sitting on a cluster of a pair of VM cluster nodes spread across a cluster pair of real nodes? It could be go beyond pairs but trying to ask the question using thousands of nodes and levels would go pear shaped and I'd confuse myself. Ah, a nice frosty beverage... or two.
Hyper Visor is Translation between NT5 & NT6. It resides in multiple places, mainly in Host & secondary Desktop/terminal units. However, Hyper visor is way to translate 16 bit to 32 bit & 32/64 bit interchange, so it may lie in router, modem or other hardware, such as HDD, in bitie needed areas to insure proper translation.
By explaining Stack of Partiton so well, Corrupted stack can cross over to anotehr partition, Guessing "What is HyperV, is new GAAGaGod. Hypervisor is interconnect snippet so machine that plays windows ?9Xp8 & its games, can be connected to system of Ultimate. Also ,large Farmers with Plenty of Good Older equipment can still continue UpGrade Path, without trashing enire present system, as opposed to Stack crash without recovery disk or permanent flaw.
vondrashek
Stack is actually order software puts down its files into HDD. also way files are used are part of stack. Partition is Build as it is built up. Failed stack can corrupt build, even destroy it. Sometimes infection can race to another partiton from collasped Stack, causing Multiple partitiion failures.
In ye oldie times Harmonic rate stack was created was important part of SPEED Dial useage,. rate today is controlled internally by hardware & software.
Signed:PHYSICIAN THOMAS STEWART von DRASHEK M.D.
One thing that reminds me, that NO Computer has SPEED Dial AcceLErator anymore, yet sometimes stack won't put down, say corrupted build won't take emergency update due to FLUTTER in system.
Heres Workaround. Back in Really Old Days, Speed dialer worker near Frequency of Sound, so puching extra 4K in Mhz/s was easy. Now of course its entire Maritan Mess, yet in critical upload or install stalls, turn on audio player or steaming event or dvd insert that racks system bit & unit might break from stack freeze & complete update install & Save 'd DAY.
Signed:PHYSICIAN THOMAS STEWART von DRASHEK M.D.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
