Intel and hypervisor provider Citrix Systems are announcing a joint collaboration that will see a bare-metal hypervisor based on the commercial variant of the Xen hypervisor delivered later this year. While there are a number of different ways that you can use hypervisors to virtualize your x86 or x64 desktop or laptop computer …
I think enough time has passed for the Innotek reference to be relevant... it's been a year since the acquisition, it might be time to accept that it's part of Sun already.
Or maybe you would like to add the MS Virtual PC came from the Connectix acquisition... or that Parallels Desktop came from the twoOStwo development paid by Netsys and might be legally dubious...
Aside from that, I don't see that as such a big deal... Xen, VBox, ESX, VMware workstation and most other hypervisors alredy make use of the hardware extensions... and if you want a true hardware hypervisor, go check Sun's sun4v arquitecture, also known as Niagara... LDOMs is truly a H/W hypervisor, monitored from S/W... and not only that, it's completely opensource, both hardware and software... how can intel/citrix compare with that?
Could be an interesting fight
A hypervisor is a piece of software that lets multiple copies of another piece of software each believe they have the entire system to themselves. Back in the 1950s, this was called an operating system. What is proposed here is a return to a micro-kernel operating system, with the desktop personality (aka, OS) running on top.
So Intel/Citrix want to make themselves the de facto standard for such things, eh? Well there are already three players in that market and as Young Mr Gates observed many years ago, the OS is a natural monopoly. Ultimately, there will only be one survivor. (My vote goes for the free offering. After all, if your hypervisor is so complex that paid-for support is important, then it's too big!)
There's also the nagging worry that however elegant the idea might be, it has never been particularly successful in practice. How many great micro-kernel OSes can you name?
I thought the sun4v architecture could only do SPARC-based OS, so maybe it's not really that comparable to the x86 stuff that this article is referring to.
Oh and Virtualbox isn't open source - it's licensed for commercial use. (Yes I do know about the PUEL)
Heck, if we're talking hypervisors, then let's talk about the one that IBM ship in their Power (System p) kit - okay it's not open source. But then again, it's based on mainframe tech and gives some pretty clever virtualization options. Sorry if I sound like an IBM advert (no, I don't work for them) but the tech is really quite impressive.
"A hypervisor is a piece of software that lets multiple copies of another piece of software each believe they have the entire system to themselves. Back in the 1950s, this was called an operating system. What is proposed here is a return to a micro-kernel operating system, with the desktop personality (aka, OS) running on top."
Good heavens - how many misconceptions. Firstly in the 1950s operating systems pretty well didn't exist. Early applications were written to the bare metal - that is they talked directly to the hardware. Gradually manufacturers introduced tools and software to make this easier to do, although the low level hardware interfacing code was still included in each application. By the 1960s operating systems had emerged which were environments in which a single application could run within a controlled environment offering some higher level services. Things such as I/O could abstracted to operating system services, although even then some applications got very close to the hardware devices by direct calling libraries that issued I/O instructions to the hardware (dig into z/OS and you will find that many of the older access methods have still got this structure, although the OS intercepts the actual I/O call). These early operating systems would only run a single applications at a time and only later did versions appear which allowed multiple applications to (apparently) run simultaneously. Early versions of these multi-taking operating systems (and the hardware on which they ran) tended to lack mechanisms to protect them from each other - a badly written application could write all over the storage belonging to another. Modern multi-tasking operating system with proper security and isolation didn't really become common until the 1970s. (PCs followed the same path, maybe 15 years later).
There is a fundamental difference in intent between an operating system and a hypervisor. Essentially the former is designed to provide services and an environment within which applications are run. In consequence, the services offered are largely abstracted ones of use to the application.
In contrast, the intent, of a hypervisor (at least in its purest sense) is to emulate a physical server. If implemented properly it will virtualise CPUs, I/O devices, memory and so on. The vast majority of business applications are not written to work in such an environment. In the case of mainframes, hypervisors started out as pure software (VM being the obvious example) and efficiency was gradually improved through moving resource-intensive virtualisation tasks into microcode. Eventually pure micro-coded hypervisors appeared (a route down which microprocessors are heading albeit using direct silicon logic and not via microcode).
There are similarities between some of the technologies used in operating systems and hypervisors, but the second is a much more constrained thing. Its job is to pretend to the code it is hosting that it is a physical machine complete with I/O ports, clocks, interrupts and so on. . VMWare ESX and the like are not throwbacks to the operating systems of the 1950s (which didn't exist in a recognisable form at that time anyway), but to VM which emerged during the 1970s. Of course there are lots of grey areas - hyperviser aware operating systems, hypervisers running within a general purpose operating system, not to mention completely abstracted virtual machines, such as the Java VM. But the principle remains.
There is a lot of common technology in hypervisors and microkernels, but the latter is really the set of low-level primitive services required to allow for the implementation of an operating system whilst a hypervisor (in its purest sense) is there to provide virtual machine environments.
So you think that most Intel chips are VT enabled
So you think that most Intel chips are VT enabled! This is just not the case on low end laptops and desktops. Now if you want low end windows compatible (guest) piece of hardware get AMD X2 they have AMD-V built in and the memory controller.
You can make a Citrix XenServerExpress 5.0 (limited 4 vms) box for 200 quid when you use AMD X2 machines.
being different architectures doesn't make them less comparable... the Intel/Citrix plan is nothing new... Intel is late to the game, as usual (just as with SMP, CMT, 64-bits CPUs and so many other things)
but as usual, the lateness of innovation from Intel is celebrated by the ignorant crowd... just like windows with MS... when will we be free of the Wintel monarchy...?
About sun4v being SPARC only... well, that's obvious, but OSs only need to be ported... just ask the Ubuntu and NetBSD people about that... OpenSparc's licence and openness is allowing the architecture advance a lot more than what it seems, I just hope it continues with that trend
Virtualbox IS open source, under the GPLv2 licence... there is another version, ruled by the PUEL, but vbox is open source, so stop spreading FUD... let's wait and see when VMware Workstation and Parallels Desktop become as open source as vbox.... protip: don't wait for it standing