Submitted by “Dell’s chief technology officer sees a huge future in Linux virtualisation for the once-dominant PC manufacturer. Kevin Kettler told an audience at LinuxWorld that virtualisation and Linux was no longer such an odd combination. “The two play to one another very strongly, particularly in the re-emerging trend of virtualisation,” Kettler said.”
Insert *virtualisation* here.
Although virtualisation is undoubtedly useful, in reality you trade off one set of problems for another.
segedendum! We agree! This is almost a touching moment.
But yes, virtualisation is most certainly useful, but also, most certainly overhyped at this point.
Misunderstood, for some yes. Interesting, for others yes. Over hyped, in some situations yes. In general however, I certainly wouldn’t offer up the comment that virtualisation is over hyped. It’s actually really useful for consolidation and cost savings.
You also have to consider what type of virtualisation you’re talking about. Take Sun Solaris’ zones concept (some call them containers). It’s virtualisation, but rather different to that of VMWare or alike.
At least for my employer (a managed infrastructure services organisation), our clients are moving towards consolidation via virtualisation because of the benefits. For us at least, it’s moved beyond hype and become reality.
When was virtualisation + linux an odd combination?
I’m more curious about when Dell became a “once-dominant PC manufacturer”. They still seem pretty dominant to me..
Virtualization allows people to run Linux without rejecting the option of running Windows.
It’s a very nice way to break the barriers imposed by The Monopoly. Virtualization is by far the nicest thing that happened to Linux in its fight against Windows.
Although virtualisation is undoubtedly useful, in reality you trade off one set of problems for another.
In all of the hype that began with some niche client solutions and then exploded into the server market, even CTOs of major IT vendors have lost sight of what virtualization actually is. Imagine how the average CIO feels…
Virtualization is the abstraction of system resources. The virtualized resources have two advantages over their backing resources: multiplicity (you can replicate a resource) and commonality (similar resources appear identical).
Yes, you can abstract the entire system, creating multiple copies of a common system architecture. This is platform virtualization, and it’s where the hype begins and largely ends in the story of virtualization.
But virtualization is everywhere. We use it to allow several filesystems to share a common namespace (virtual filesystem). We use it to map alternative and/or imaginary processors onto our native processors (virtual machine). We use it divide system memory into isolated address spaces (virtual memory).
Virtualization isn’t just a relic from the mainframe era that’s coming back in style. It isn’t the new black. It’s always been THE black. Computer science is essentially the study of problem solving through a network of abstractions. From abstracting whole platforms with Xen on up to abstracting data with linked lists, it’s all the same underlying idea.
So it stands to question if virtualization is over-hyped or under-hyped? No, it’s just mis-hyped.
Linux and virtualization is no longer such an odd combination? It’s interesting how great ideas stop being odd when leaders stop being ignorant. If flexibility were as important to selling computers 10 or 20 years ago as it is today, then virtual platforms and virtual operating systems would have been right up there with virtual memory and virtual filesystems.
Why is flexibility so important to selling computers today? Because of Linux.
Edited 2007-08-13 19:03
Yeah sure linux is the reason to why flexibility is so important today.. The fact that it’s easier then ever to find similar products on the internet from another providor has nothing to do with it.
I mean yer, virtualisation is useful but there are just so many people who think that it will somehow revolutionise the OS. That first post was spartan on details.
If you’ve ever looked at your server room, seen the unused servers and capacity and thought “What the hell do we need all that stuff for?” then virtualisation is one of your best bets. It can save you a lot of money if you play your cards right.
The first problem that you have is that if you consolidate machines on to a smaller number of physical machines then the obvious thing that you give up is hardware redundancy. You better hope that your hardware is reliable and the underlying system that you run your hypervisor on is sold as well, otherwise it all unravels. Many solutions have tried to build in redundancy by sharing virtual machines between different physical machines, but it’s just more complexity for the sake of solving a problem you wouldn’t otherwise have. You’d be better off skipping virtualisation and just having separate physical machines.
Software installation is often muted as an advantage of virtualisation, and it is – in a fashion. You get to have a base VM image with what you want installed on it, and you deploy that. However, you have to specify and decide what goes on it and only you can truly customise it to your needs and reuse it for your own use. A lot of these virtualisation people are jumping up and down about so called virtual appliances, and how you’ll not have to do any software installation. Not true at all.
We’ve even got people now claiming that hypervisors will replace operating system driver by having all the drivers in the hypervisor! Some of this virtualisation stuff from some people is insane.
It’s like a guy waking up one day and finally discovering his penis. It’s been there all along, but it’s very new and exciting nonetheless. There are many places to put it, some more appropriate than others, and successful deployment requires a bit of subtlety and good judgment. It will be the cause of erratic behavior and embarrassing moments as he learns to harness its potential.
But that doesn’t make it any less central to our fundamental objective.
Behold, an analogy that only vaguely involves cars.
Butters,
I have a huge respect for you. I absolutely adore most every post you make; I look for them.
I’ve noticed you moving from purely technical posts into the area of presenting analogies.
And this one is absolutely A-1 On-The-Mark.
You left out the bit about it having a huge down-side potential, and it being potentially fatal, though.
True enough! Virtualisation adds some complexity and also doesn’t solve the most fundamental problem in IT nowadays: how to manage so many servers at the same time?
Server provisoning, configuration management, availability, flexibility and so on, that’s the real problem, not the money you spend on hardware nor physical infrastructure (even though electricity bill are raising quite fast lately in the datacenters). Sun got it right there when they moved all their product strategy toward providing a flexible “Operating Environment”.
Also Kettler’s demo is the classical marketing stuff, full of non-sense and technical inanities. What the heck does he means by “embedding the hypervisor”? Doing some kind of blackbox appliance such vmware esx server? Then since he made all his demo with Xen which is useless without a linux dom0, they will compete with Vmware but also XenSource and all the other linux distro since they all have a similar product. What’s the interest for Dell? Selling Xen-certified servers?
Honestly, even though Xen was an interesting hack back in the time when HVM did not exist and it required the host OS to be modified. But since Dell is talking about running unmodified OS, they’ll require HVM along with Xen.
And on that segment (full virtualisation with HVM) Xen is a bloated hack compared to KVM (kvm.qumranet.com). KVM is technically so superior to Xen microkernel approach that it’s already included in the linux kernel and that its development has nearly overtook Xen in 1 year time even though less people were working on it at the time. Since then the kernel hackers interest has raised and it now includes stuff such a swap for the host allowing memory overcommitting that Xen doesn’t have (and probably never will as this is a design limitation afaik).
So even amongst the full machine virtualisation space, Dell went for the wrong partner. Instead they should be contributing to KVM or even better helping the join effort of the container guys (Eric Bierdmann, Paul Menage, devs from vserver, openvz and IBM/Meiosys) who aim at providing lightweight system virtualisation, freeze/migrate habilities for the processes and so on.
I had never seen anyone do production virtualisation until I saw it on linux and unix??